Skip to content

Sync upstream v1.4.0 (v1.4.0+mas.1) - #4

Merged
mbarrerarod merged 1463 commits into
mainfrom
cursor/sync-upstream-v1.4.0-cdbb
Nov 12, 2026
Merged

mbarrerarod merged 1463 commits into
mainfrom
cursor/sync-upstream-v1.4.0-cdbb

Conversation

@mbarrerarod

Copy link
Copy Markdown

Description

Syncs MAS-Consultores/spec-kit with upstream github/spec-kit v1.4.0, preserving all MAS customizations.

Upstream Changes Merged (v0.8.10 → v1.4.0)

  • Modular command architecture (init refactored to command_init.py)
  • New integrations added (forge, goose, kiro-cli, etc.)
  • Enhanced preset system with catalog support
  • Improved shared infrastructure management
  • New --extension flag for init
  • Constitution sync preset
  • Various bug fixes and improvements

MAS Customizations Preserved

  • Stack-aware initialization: --stack parameter required (cakephp2-mysql, moodle3, moodle5-plugin, moodle5-portal, laravel-inertia-react)
  • Security guidelines: Per-stack security profiles and controls
  • Language policy: Spanish-first user-facing content with --language-policy options
  • MAS presets: mas-core + mas-stack-<id> bundled preset composition
  • Constitution: MAS-specific organization constitution template
  • Stack memory files: stack.md, security-guidelines.md, stack-context.md, language-policy.md
  • External skills: Installation from stack-skills.yml manifests

Conflict Resolution Summary

File Resolution
pyproject.toml Merged MAS presets with upstream additions; version set to 1.4.0+mas.1
src/specify_cli/__init__.py Adopted upstream modular structure with MAS imports
src/specify_cli/command_init.py Ported MAS stack-aware init to upstream's refactored module

Version

1.4.0+mas.1

(PEP 440 local version format)

E2E Stack Test Results

All 5 MAS stacks verified working via end-to-end CLI tests:

Stack Status Notes
cakephp2-mysql ✅ PASS Stack files created, speckit commands generated
moodle3 ✅ PASS Exception track properly configured
moodle5-plugin ✅ PASS Standard track working
moodle5-portal ✅ PASS Standard track working
laravel-inertia-react ✅ PASS Gemini TOML format verified
Invalid Stack Test ✅ PASS Clear error message shown

Test Suite Status

  • 94 critical tests pass (20 MAS + 74 CLI tests)
  • 4 tests skipped (upstream --preset tests, mutually exclusive with MAS --stack requirement)
  • Some upstream integration tests need --stack parameter added (expected for fork with mandatory parameter)

Testing

  • Tested locally with specify --help
  • Ran MAS-specific tests with pytest tests/test_mas_stack_init.py tests/integrations/test_cli.py -v
  • E2E tested all 5 stacks with specify init <tmpdir> --stack <stack> --integration <agent>
  • Verified stack memory files are created
  • Verified speckit-* command templates are generated
  • Verified invalid stack gives clear error

AI Disclosure

  • I did use AI assistance

AI disclosure: Claude Opus 4.5 via Cursor Cloud Agent (autonomous mode, high reasoning effort). Used for merge conflict resolution, code porting from MAS's old init.py to upstream's refactored command_init.py, test updates, and E2E verification.


Notes

⚠️ Do NOT merge yet - awaiting review

⚠️ MAS Release Notes Correction: The release notes at docs/releases/notes/v1.4.0+mas.1.md (if created) should reference git+https://github.com/MAS-Consultores/spec-kit.git for installation, not github/spec-kit.

⚠️ No tag/release created - per instructions, only PR opened

Open in Web Open in Cursor 

mnriem and others added 30 commits October 9, 2026 10:15
* feat: automate community workflow step submissions

Add a maintainer-triggered issue submission flow with metadata-only validation, tag-pinned file digests, scoped draft catalog PRs, and contributor documentation.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: harden workflow step submission validation

Keep setup failures recoverable, reuse CLI step ID validation, and gate per-file downloads and hashing through a tested repository-owned helper.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: enforce installer limits for submitted step packages

Validate file and directory counts and nesting before downloading, fetch complete packages with a cumulative byte budget, and cover rejected package paths and exact limit boundaries.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: classify transient HTTP errors before curl size failures

Prioritize known HTTP responses over curl exit codes, treat HTTP 408 as a blocked timeout, and cover status/exit combinations with ordinary and one-byte remaining budgets.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: reject case-insensitive step package path collisions

Compare package file paths component-wise with casefold before downloads, rejecting duplicate file aliases and file/directory conflicts while preserving noncolliding paths.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* docs: clarify workflow step submission activation and provenance

Document maintainer-label activation and distinguish the required archive download URL from catalog per-file distribution URLs.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: align workflow step archive and documentation contracts

Reject unsupported GitHub-generated tgz archives while retaining named tgz release assets, and accept any tag-pinned Markdown usage document rather than requiring README.md.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: clarify step license validation and author attribution

Require valid SPDX identifiers and matching release-license metadata before catalog edits. Align table rendering with other community submissions without conflating package authors and maintainers; preserve Decision maintainer attribution in prose.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: enable public REST access for tagged step license metadata

Allow api.github.com in agent URL and firewall policies and prescribe the enabled web-fetch route. Cover compiled access, isolation, and retained host restrictions with a regression test.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: support epoch releases and document submission URL alphabets

Accept PEP 440 epoch markers in release tags and archive URLs while retaining package path restrictions. Document accepted character sets and percent-escape exclusions, and cover epoch route compatibility and mismatches.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: isolate step downloads from curl config and proxy auth failures

Place --disable first in curl arguments, verify hostile config isolation with real curl and a local proxy, and classify HTTP 407 as an environment blocker before exit-code handling.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* test: compare license REST allowlist domains exactly

Replace ambiguous hostname membership assertions flagged by CodeQL with explicit exact-entry comparisons and add lookalike-domain negative cases for firewall and ingestion policies.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: remove undisclosed step documentation policy requirement

Keep discovery-only policy in Spec Kit guidance without rejecting complete external usage documentation for not repeating it.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: verify step catalog history and shared submission contracts

Follow the preset verifier snapshot/generated pattern to enforce release migration, preserved history, complete existing digests, and publication gating. Include workflow-step in common community contract coverage while retaining archive-specific matrices.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: preserve catalog reader error classification by phase

Align with the preset verifier by honoring the caller's error type for filesystem and decoding failures. Cover missing, non-UTF-8, directory, and permission failures without changing original-evidence blocker semantics.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: reject duplicate JSON fields in step submission evidence

Detect duplicate object keys at every depth before snapshot comparison, preserving generated repair, original-evidence blocker, and submission-failure classifications.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: certify canonical metadata before step catalog snapshots

Compare every required current-entry metadata field against independently captured form values, including archive and repository provenance, compatibility, and optional changelog. Reject incomplete or miscopied entries before generating expected catalog state.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Add epic extension submitted by @pmckeown to:
- extensions/catalog.community.json (alphabetical order)
- docs/community/extensions.md community extensions table

Closes github#4891

Assisted-by: GitHub Copilot (model: gpt-5.2-codex, autonomous)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Update intake-review-governance preset submitted by @hindermath:

- presets/catalog.community.json (version, download_url, digest, documentation)

- docs/community/presets.md community presets table

Closes github#4889

Assisted-by: GitHub Copilot (model: gpt-5.2-codex, autonomous)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…ithub#4895)

* Update Applied Epistemic Engineering extension to v1.2.0

Update aee extension submitted by @tbitcs:\n- extensions/catalog.community.json (version, download_url, checksum, dependencies, commands)\n- docs/community/extensions.md community extensions table (no row changes required)\n\nCloses github#4890\n\nAssisted-by: GitHub Copilot (model: gpt-5.2-codex, autonomous)\nCo-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Add AEE optional tools to catalog

Co-authored-by: mnriem <15701806+mnriem@users.noreply.github.com>

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: mnriem <15701806+mnriem@users.noreply.github.com>
…hub#4764)

* feat(workflows): compose workflows with a unified execution tree

Keep invocation results and workflow bindings on the same execution occurrence. Isolate fan-out contexts and resume persisted expansions through one executor.

Assisted-by: OpenCode (model: gpt-6-astra, autonomous)

* fix(workflows): complete composition cleanup phase one

Assisted-by: OpenCode (model: gpt-5.6-terra, autonomous)

* refactor(workflows): centralize execution tree access

Assisted-by: OpenCode (model: gpt-5.6-terra, autonomous)

* fix(workflows): narrow composition call failures

Assisted-by: OpenCode (model: gpt-5.6-terra, autonomous)

* fix(workflows): replay execution projections

Assisted-by: OpenCode (model: gpt-5.6-terra, autonomous)

* fix(workflows): qualify execution events

Assisted-by: OpenCode (model: gpt-5.6-terra, autonomous)

* fix(workflows): share execution snapshots

Assisted-by: OpenCode (model: gpt-5.6-terra, autonomous)

* docs(workflows): document composition resume

Assisted-by: OpenCode (model: gpt-5.6-terra, autonomous)

* fix(workflows): preserve replay projections

Assisted-by: OpenCode (model: gpt-5.6-terra, autonomous)

* fix(workflows): reject inconsistent execution trees

Assisted-by: OpenCode (model: deepseek-v4.1-flash, autonomous)

* fix(workflows): narrow workflow target resolution failures

Assisted-by: OpenCode (model: deepseek-v4.1-flash, autonomous)

* refactor(workflows): replay fan-out through the live path

Assisted-by: OpenCode (model: deepseek-v4.1-flash, autonomous)

* fix(workflows): match main events for unknown step types

Assisted-by: OpenCode (model: deepseek-v4.1-flash, autonomous)

* fix(workflows): omit unprojected fan-out item results

Preserve the item traversal's projection decision for missing step types. Cover sequential and parallel execution, inherited aliases, unnamed templates, and resume after reinstalling the implementation.

Assisted-by: OpenCode (model: gpt-6-astra, autonomous)

* fix(workflows): preserve composed fan-out aliases

Keep fan-out item aliases in the enclosing workflow context without exposing child workflow internals in root results. Normalize unnamed templates to the item ID for result aliases and lifecycle events, and recognize generated item aliases during fan-in validation.

Assisted-by: OpenCode (model: gpt-5.6-terra, autonomous)

* Refactor condition for raising ValueError on execution

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* fix(workflows): validate composition checkpoint boundaries

Assisted-by: OpenCode (model: gpt-5.6-terra, autonomous)

* fix(workflows): make fan-out item aliases reporting-only

Fan-out item aliases (fan:template:N) are not valid fan-in wait_for targets.
In a loop iteration >= 1 the authored alias is never refreshed, so a fan-in
that waited on it silently joined the previous iteration's values. Joins now
use the fan-out step ID and its ordered output.results instead.

Revert the fan_out_aliases validation allowance so wait_for again requires
declared step IDs (matching main), and reject ':' wait_for entries at runtime
in FanInStep.execute() so unvalidated runs fail loudly rather than joining a
stale or empty item result.

Assisted-by: OpenCode (model: deepseek-v4.1-flash, autonomous)

* fix(workflows): stop logging after checkpoint failure; treat unknown child steps as call failures

RunState.append_log() now checks _checkpoint_failed under the state lock
(lock order _lock -> _log_lock), so no concurrent fan-out worker can append
a completion or nested start event after another worker's checkpoint failed.
This restores the documented guarantee that a checkpoint failure prevents
further writes.

Remove the error-text classification of unavailable child implementations at
the workflow-call boundary. A missing implementation stays terminal where the
step occurs (root and fan-out, as on main); at a workflow call it is a reported
child failure, whether detected at bind time or after binding, so call-level
continue_on_error treats both consistently and step-controlled error strings
are no longer interpreted.

Assisted-by: GitHub Copilot (model: claude-opus-5.5, autonomous)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* test(workflows): cover resume snapshot check with native YAML scalars

Add a resume regression for a workflow snapshot containing `.nan`
(max_concurrency sequential fallback) and reordered keys, and cover the
previously untested rejection of a changed workflow snapshot. Both pass on
the existing comparison: PyYAML decodes `.nan` to a shared object, so the
snapshots compare equal without a production change.

Assisted-by: GitHub Copilot (model: claude-opus-5.5, autonomous)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Refactor executor run with error handling

Refactor execution handling to ensure active step is updated even if an error occurs during execution.

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* fix(workflows): report qualified current_step_id; keep typed gate messages; halt fan-out on worker exceptions

current_step_id again uses the occurrence ID that events and step_results
use, as on main: a paused or failed root fan-out item reports `fan:item:0`
and a later loop iteration `loop:body:1`. Inside a workflow call, the ID is
relative to the called workflow. walk_execution() now derives this ID with
the executor's qualification rules, so the post-run reconciliation (kept for
exception paths) and the gate payload's step_id report the same value.

Gate `message` keeps its typed template result (for example 42 stays an
int). Only values the JSON checkpoint cannot persist, such as native YAML
dates, are stored as text; presentation still stringifies.

A fan-out worker exception now sets the halt flag before re-raising, so
earlier items finishing in order no longer dispatch further items after a
later item has already raised.

Assisted-by: GitHub Copilot CLI (model: claude-opus-5.5, autonomous)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix(workflows): persist the active step before it executes

The executor updated current_step_id and current_step_index only in memory
and wrote them with the next checkpoint, after the step had finished. While a
long step ran, and after a hard crash, state.json still named the previous
step, so `workflow status` and the run payload reported the wrong occurrence.
main saves before each step starts.

Checkpoint the active occurrence before it executes, so root steps, later
loop iterations, and fan-out items are persisted under their qualified IDs
while they run. This adds one state write per started step, as on main; the
checkpoint-count tests are updated accordingly.

Assisted-by: OpenCode (model: claude-opus-5.5, autonomous)

* fix(workflows): persist workflow calls before announcing them

1b0a3a4 checkpointed the active occurrence before step_started, but only
on the shared step path. Workflow calls return to Execution.workflow()
before that path and emitted step_started without a checkpoint. Until
binding was saved, and on resume of an already-bound call until the first
child step, state.json named the previous step, and a crash or checkpoint
failure could leave a start event for the call that no checkpoint
reflected.

Move the start sequence (set current_step_id, checkpoint, then log and
callback) into one Execution.start() helper used by both paths, so the
rule lives in one place. The ID update and checkpoint now happen under one
lock hold. Tests cover top-level, nested, and resumed bound calls, and a
failed start checkpoint that must leave no start event; all three failed
before this change.

Assisted-by: OpenCode (model: claude-opus-5.5, autonomous)

* fix(workflows): let interrupts during checkpoints pause the run

RunState.save() caught BaseException around the checkpoint write, so a
KeyboardInterrupt during any checkpoint became CheckpointError, marked the
state as failed to checkpoint, and bypassed the KeyboardInterrupt handler
in execute()/resume(). The run stayed "running" and could not be resumed.
On main the interrupt reaches that handler and the run pauses. Introduced
in 7aa424d.

Convert only Exception to CheckpointError. Write failures still stop
further writes; interrupts propagate unchanged, and the pause re-saves the
tree from memory. Tests interrupt a checkpoint at the root, inside a
workflow call, and during resume; all failed before this change.

Assisted-by: OpenCode (model: claude-opus-5.5, autonomous)

* fix(workflows): replay the result of aborted fan-out items

6792bea made fan-out items publish only the result their traversal
projected, so missing step implementations publish nothing. The replay of
an already-aborted occurrence returned without projecting its stored
result, so after a sibling item resumed, the aborted item appeared as {}
in the fan-out's ordered results.

Project the stored result before returning "aborted", as the live run did
when the occurrence aborted. Nothing is re-executed and no events are
emitted. The aborted-sibling regression now checks the item output for a
step, an if container, and a workflow call template; all three failed
before this change.

Assisted-by: OpenCode (model: claude-opus-5.5, autonomous)

* fix(workflows): hide partial fan-out results from resumed items

A halted fan-out checkpoints its partial results in its record for
reporting. On resume that record is projected before the items restart, so
re-executed and not-yet-started items saw steps.<fan>.output.results with
a stale prefix, including the paused item's own earlier output. Items in a
live run, and on main, never see results: the engine adds them only after
the items finish. Present since 7aa424d.

Remove the engine-added results from the fan-out's own entry in the
item-local snapshot. The parent context and checkpoint keep them.

Add an invariant test: a resumed item must see exactly the context it has
in a run that never paused, for step, if-container, and workflow-call
templates, sequentially and in parallel. It found only this difference;
the four affected cases failed before this change.

Assisted-by: OpenCode (model: claude-opus-5.5, autonomous)

* fix(workflows): validate occurrence lifecycle and report interrupted scopes

Persist fan-out templates as YAML, reject invalid resume offsets and tree states, and check only changed nodes during lifecycle transitions. Preserve original step exceptions during best-effort cleanup and derive interrupted bound-call status from the run. Cover regressions and document the contract.

Assisted-by: OpenCode (model: gpt-6-sol, autonomous)

* test(workflows): exercise legacy migration checkpoint failure

Assisted-by: GitHub Copilot (model: GPT-5.6 Luna, autonomous)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix(workflows): discard a resumed call's previous attempt status

A workflow call that paused kept its recorded `status: paused` result
while blocked. When resume re-entered the call and a later child step
raised, the call stayed blocked with that stale result, so status
reported the scope as paused even though the run had failed.

Re-entering a blocked, bound call now returns it to its children phase
and drops the previous attempt's result, outcome, and error. Its
children hold the authoritative progress; an exception leaves no
recorded call status, so the scope reflects the run status. Leaf steps
keep their result because finish replaces it on re-execution.

Assisted-by: GitHub Copilot (model: Claude Opus 5.5, autonomous)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* docs(workflows): describe current_step_id as a compatibility view

The public reference and the architecture document promised that
`current_step_id` always identifies the active execution-tree leaf. It
does not: it is updated as occurrences are entered, but completion does
not reconcile it, so a parallel fan-out can leave it naming an item that
has already finished while another item remains active.

Apply the reviewer's suggested wording: document `current_step_id` as a
compatibility occurrence-ID view rather than an exact active-step set, and
identify the tree's per-occurrence `active` flags as authoritative.

Assisted-by: OpenCode (model: deepseek-v4.1-flash, autonomous)

* fix(workflows): discard stale status of every re-entered blocked call

A workflow call whose binding failed (for example, an uninstalled
target) is blocked with a failed result but no binding. The re-entry
cleanup from 53a434d only ran for bound calls, so a retry that then
bound successfully carried the old result into the children phase. If
the child was interrupted, status reported the scope as failed while
the run was paused; after an exception, the call kept the stale binding
error.

A call's recorded result summarizes its previous attempt, so begin now
drops result, outcome, and error for every blocked workflow call. A
bound call continues at its children; an unbound call returns to ready
and binds again. Leaf steps keep their record until finish replaces it,
because it describes that same step: an interrupted resumed gate still
reports its prompt in status, as on main.

The unbound-call regression fails before this change (scope reported
failed instead of paused; stale error retained). The gate test guards
against extending the reset to leaf steps.

Assisted-by: OpenCode (model: claude-opus-5.5, autonomous)

* docs(workflows): state interrupt handling of the setup checkpoint

The RunState.save() comment said a KeyboardInterrupt always reaches the
pause path of execute()/resume(). That holds during step execution, but
the setup save before execution is outside that handler: an interrupt
after it replaces state.json leaves the run `running`. Describe that
accurately.

Also drop comparisons with the main branch from code comments, a test
comment, and the architecture document; once merged they no longer
describe anything. No behavior change.

Assisted-by: OpenCode (model: claude-opus-5.5, autonomous)

* fix(workflows): stop gate interrupt test leaking into later tests

test_interrupted_resumed_gate_keeps_its_prompt_in_status patched execute
on the shared STEP_REGISTRY["gate"] instance. Undoing that patch leaves the
original bound method as an instance attribute, which shadows later
class-level patches of GateStep.execute. In the full suite this made
test_resume_preserves_original_workflow_dir stay paused after resume.

Patch the GateStep class instead so teardown restores it cleanly.

Assisted-by: OpenCode (model: claude-opus-5.5, autonomous)

* test(workflows): pin and document the fan-out step_template removal

The rewrite stopped publishing a fan-out's raw `step_template` in its
step output, but nothing checked it end to end: the only `step_template`
assertions call FanOutStep.execute() directly, before the engine strips
the field. The public workflow reference did not mention it either.

Add an engine-level test that runs a fan-out, pauses, and resumes. It
checks that the template stays in the private execution tree and is
absent from the live, replayed, and persisted fan-out output, from a
fan-in aggregate, and from expressions. It fails when the engine
publishes the template again.

Add a migration note to docs/reference/workflows.md describing the
removed field and where to read the template instead.

Assisted-by: OpenCode (model: claude-opus-5.5, autonomous)

* docs(workflows): clarify rebind failure recovery

Document that retrying a previously bound workflow call discards its stale result, outcome, and error while preserving the binding and child subtree. Also describe the ready state used when an initial binding failed.\n\nAssisted-by: GitHub Copilot (model: GPT-5.6 Luna, autonomous)\nCo-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Markus <markus@example.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Markus Wondrak <markuswondrak@gmail.com>
…ithub#4844)

* feat(bundles): reconcile exact component pins and version conflicts

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Fix selected bundle release installation and catalog outage validation

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Fix catalog validation precedence and selected workflow paths

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Fix catalog redirect and encoding validation

Reject unsafe extension and preset catalog redirects during exact-ID lookups and classify malformed extension encodings. Cover stacked and legacy fetch paths with before-and-after regressions.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Validate deeply nested workflow and step catalogs

Classify JSON recursion during parsing or cache writes as malformed catalog data, recover from poisoned caches, and keep non-fetch step catalog errors blocking exact-ID lookup.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Align bundled preset validation with installation

Reject unsourced bundled preset pins that the installer cannot satisfy, classify extension HTTP protocol failures as unreachable catalogs, and make deep-cache regressions deterministic across Python versions.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Reject missing catalog collections and bundled extension pin mismatches

Require workflow and step collections in fetched or cached catalogs before exact-ID lookup may fall through. Align bundle validation with bundled-first extension installs while retaining explicit catalog sources and matching installed versions.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Reject duplicate IDs in list-form workflow catalogs

Treat duplicate workflow IDs within one list-form catalog as invalid during exact-ID lookup and search, while retaining priority-based resolution across distinct sources.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Report partial catalog outages during targeted lookup

Preserve uncertainty when an exact bundle reference is absent from readable catalogs but another configured catalog could not be fetched. Retain winning-source and untargeted search behavior; cover all four catalog families and CLI validation.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Persist all installed bundle requirements for conflict checks

Keep independently installed component pins separate from bundle-owned contributions. Reject incompatible later installs, retain requirements on remove, and preserve legacy records' known pins.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Protect other bundles' pins during unpinned installs

Preserve ownership attribution while consulting all recorded pinned requirements in preflight. Reject unpinned installs or refreshes that may replace a required version and allow no-op sharing of a matching installation.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Describe bundle overlaps as shared requirements

Avoid attributing ownership when another bundle only requires an independently installed component. Cover conflict reports and bundle info output for required-only records.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Reject lower-priority catalog matches after source outages

Preserve precedence for exact component lookups when an earlier catalog could not be fetched. Keep broad search available and warn when bundle validation cannot verify the winning source.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Protect shared bundle install requirements

Reject sharing or refreshing a component when another bundle requires different source or preset priority/strategy. Check required-only records before any primitive mutation and retain compatible sharing.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Transfer ownership when bundle contributions remain needed

Keep bundle-installed components attributed after removal or update drops the sole contributor while another bundle still requires them. Preserve independent ownership and cover sequential removal and update cleanup.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Classify malformed extension and step catalogs consistently

Treat deeply nested extension JSON as catalog validation failures across network and cache paths, and classify duplicate step IDs as malformed catalog data. Cover targeted and discovery lookups with regressions.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Reject non-string IDs in workflow catalogs

Classify malformed list-form workflow IDs as catalog validation failures before duplicate detection or exact-version resolution. Cover targeted and search lookups for unhashable, numeric, boolean, and null IDs.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Reject ambiguous bundle sources and malformed step IDs

Fail closed when an explicit bundle component source names multiple active catalogs, including validation and direct primitive installation. Reject non-string list-form step IDs and update the former coercion regression. Simulate JSON parser recursion deterministically across Python versions so CI tests the exception contract rather than interpreter-specific depth limits.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Count required components in bundle list output

Report the complete required component set even when every component was already installed independently. Preserve legacy record fallback and JSON provenance, document unversioned catalog pin behavior, and make unknown-step reference testing independent of network availability.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix(bundle): confine exact-pin resolution to bundler

Keep strict catalog precedence and selected historical releases within bundle validation and installation. Restore general catalog behavior to upstream main while preserving workflow and step preselected installer hooks.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix(bundle): validate pinned current release metadata

Reject malformed workflow and step install URLs and checksums during online bundle validation, including the advertised current release. Preserve legacy workflow digest compatibility and cover both valid and invalid pins.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix(bundle): reject unsafe pinned extension URLs

Validate selected current and historical extension release URLs during online bundle checks so invalid references fail before installation. Add regression coverage and keep catalog fixtures installable.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix(bundle): validate pinned current component metadata

Reject invalid current preset URLs and malformed declared extension/preset digests during online exact-pin validation. Keep omitted digests and unversioned legacy entries compatible; cover failing and valid cases.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix(bundle): reject conflicting catalog entry IDs

Reject declared IDs that disagree with mapping keys before bundle release selection. Cover all four component catalogs, preserve matching or omitted IDs, and order test imports consistently.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Update intake-sequencing-governance preset submitted by @hindermath:\n- presets/catalog.community.json (version, download_url, sha256, documentation)\n- docs/community/presets.md community presets table\n\nCloses github#4899\n\nAssisted-by: GitHub Copilot (model: gpt-5.2-codex, autonomous)\nCo-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
* chore: bump version to 1.1.3

* chore: begin 1.1.4.dev0 development

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Update security-governance preset submitted by @hindermath:\n- presets/catalog.community.json (version, download URL, digest, metadata)\n- docs/community/presets.md community presets table\n\nCloses github#4901\n\nAssisted-by: GitHub Copilot (model: gpt-5.2-codex, autonomous)\nCo-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Update architecture-governance preset submitted by @hindermath:\n- presets/catalog.community.json (version, download_url, digest, metadata)\n- docs/community/presets.md community presets table\n\nCloses github#4906\n\nCo-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
* feat: add maintainer-triggered PR description assessment

Port the complete pr-assess workflow with concise reviewer-facing comments,
bounded outcome-label updates, focused tests, and usage guidance.

Keep the reviewed gh-aw v0.89.21 runtime pin isolated from existing workflows.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 9ee0ab16-f074-4303-82b9-d11bfad16175

* fix: replace pr-assess outcomes without partial cleanup

Port the tested built-in label replacement and standalone-comment behavior.
Keep matching, conflicting, or unreadable outcome labels unchanged.
Limit suggested updates to the PR description, not changes to the code.
Include offline digest-checked probes for the pinned MIT-licensed handler.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 9ee0ab16-f074-4303-82b9-d11bfad16175

* Check for Node.js availability in tests

Skip test if Node.js is not available.

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* fix: simplify pr-assess outcome labels

Follow the extension-submission remove/add pattern: remove up to two stale
outcomes and add the selected outcome only when absent.
Keep matching outcomes unchanged, post fresh standalone comments, and
limit suggested updates to the description.

Remove the obsolete replacement-handler tests and fixtures. Make no
transactional or concurrent-manual-edit guarantee.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 9ee0ab16-f074-4303-82b9-d11bfad16175

* fix: include PR title in assessment stability check

Compare title text with the existing captured inputs before reporting.
Require an inconclusive explanation when the title changes during assessment.
Update the existing prompt contract and regenerate its pinned workflow lock.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 9ee0ab16-f074-4303-82b9-d11bfad16175

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Copilot-Session: 9ee0ab16-f074-4303-82b9-d11bfad16175
* feat: add scoped FileHelper foundation

Add filesystem-owning primitives with explicit symlink policy, scoped containment, atomic target updates, recursive deletion preflight, and Windows path/reparse rejection.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: e8758245-6031-47dd-8cd9-4ed60611181e

* fix: normalize Windows extended FileHelper targets

Reuse the existing drive/UNC prefix normalization for root comparisons while retaining accessed hierarchy evidence and rejecting escaping or unsupported targets. Add portable regression and native Windows operation cases.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: e8758245-6031-47dd-8cd9-4ed60611181e

* fix: scope parent traversal checks below trusted roots

Derive operation-relative parts before rejecting parent traversal so paths built from trusted roots containing parent components remain usable. Preserve lexical root evidence and reject traversal in operation suffixes; add positive and negative regression coverage.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: e8758245-6031-47dd-8cd9-4ed60611181e

* fix: handle mixed-case Windows UNC namespaces

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: e8758245-6031-47dd-8cd9-4ed60611181e

* fix: reject existing leaves before exclusive file creation

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: e8758245-6031-47dd-8cd9-4ed60611181e

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: e8758245-6031-47dd-8cd9-4ed60611181e
* Add built-in GitHub workflow step

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

* Preserve GitHub artifact bytes across fresh workflow runs

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

* Clarify Actions bot identity constraint

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

* Address GitHub workflow step review findings

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

* Support GitHub workflow steps in fan-out

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

* Limit built-in GitHub workflow step to explicit labels

Remove comment, artifact, and PR checkout operations from the first release. Keep retries idempotent by checking existing issue or PR labels before adding the requested label, and document the smaller contract.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

* Test supported GitHub workflow origin formats

Cover HTTPS and SSH remotes with and without .git, including the repository identifier and derived API URL used by the label step.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

* Accept standard GitHub SSH origin URLs

Recognize ssh://git@github.com remotes and case-insensitive GitHub hostnames while keeping the API fixed to api.github.com. Cover allowed formats and lookalike-host rejections, and distinguish Git remotes from REST endpoints in the reference.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

* Select GitHub label repository explicitly

Require a validated owner/repo workflow field instead of reading a Git origin. Keep label operations fixed to the GitHub API and cover operation from directories without a checkout.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

* Keep bundled extension outside GitHub step PR

Restore the opt-in extension README to the target branch version. The built-in label step is documented in the workflow reference, so this PR does not ship an extension change or require an extension version bump.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

* Cover GitHub step transport failures

Exercise missing gh, timeout, nonzero exit with sensitive stderr, and malformed GET/POST responses through the step boundary without live GitHub requests.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

* Set JSON content type for GitHub label requests

Send an explicit application/json header with gh API label POST bodies while leaving GET requests unchanged. Mock the request boundary and cover the regression.

Assisted-by: GitHub Copilot (model: GPT-6 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: ec94c538-a8b2-4c39-81fc-3ba5f0a59ce6
* feat(claude): add --no-model-invocation integration option

Claude skills are always written with `disable-model-invocation: false`,
so every spec-kit skill description sits in Claude Code's context and
counts against the skill-listing budget. In projects with many skills
that pushes other skills out of the listing (github#2654).

`specify init --integration claude --integration-options="--no-model-invocation"`
now writes `disable-model-invocation: true` instead. The default is
unchanged. The option persists in .specify/integration.json like other
integration options, so `specify integration upgrade claude` reapplies
it; `--integration-options=""` clears it.

The change is confined to the Claude integration: setup() records the
option on the integration and post_process_skill_content() reads it, so
preset and extension skills re-registered in the same run follow it.
Commands that write those skills without calling setup() (extension
add, preset add, integration use) keep the default; the docs say so and
point at `integration upgrade claude` to reapply.

Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NRjt1UC5M2USX77TzRMemU

* fix(claude): keep extension skills model-invocable; document gaps

Mandatory extension hooks tell the model to invoke the hook skill, which
Claude Code refuses for `disable-model-invocation: true`. Limit
--no-model-invocation to the core command skills (and preset overrides
of them) by recording the core template stems in setup(), and reset that
state if setup() fails so it cannot leak into later post-processing.

Document the behaviour in its own section: extension skills stay
invocable, upgrade options replace the stored set, and commands that
rewrite skills without reinstalling Claude (preset add/remove,
integration use) need `integration upgrade claude --force` afterwards.

Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NRjt1UC5M2USX77TzRMemU

* refactor(claude): derive core skills from CORE_COMMAND_NAMES

Store a plain flag from setup() and decide per skill in
post_process_skill_content() using the shared CORE_COMMAND_NAMES, so
there is one definition of "core command" and no extra template listing.
Require the `speckit-` prefix so an unrelated skill named like a core
stem (e.g. `tasks`) is not affected. Say "core command skills" in the
option help.

Docs: list every command that rewrites the skills without reinstalling
Claude, and note that re-running `specify init` stores only the options
passed to it. Tests: reset the in-memory flag before upgrade in the
extension-hook test so it proves persistence, and stop asserting the
documented preset gap as required behaviour.

Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NRjt1UC5M2USX77TzRMemU

* fix(claude): tighten --no-model-invocation edge cases and docs

Parse the skill name once per post-processing call, and enable the
option only for a stored value of exactly `true`.

Docs: say the option is applied whenever Claude's skills are installed
or upgraded rather than promising it for every preset override, add the
`specify bundle` preset paths, limit the `switch` caveat to a Claude
install that is not the default, and warn that a mandatory extension hook
on a core command cannot run with the option on. Test the documented
preset gap explicitly so a behaviour change forces a docs update.

Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NRjt1UC5M2USX77TzRMemU

* docs(claude): narrow --no-model-invocation caveats to what reverts

Only preset-related skills are rewritten by preset, use and switch
commands, and `upgrade --force` re-registers presets only while Claude is
the default integration; say so, and clarify that the hook caveat is
about hooks whose command is a core command. Pin the documented need for
`--force` with a test, and make `_is_core_command_skill` a staticmethod.

Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NRjt1UC5M2USX77TzRMemU

* docs(claude): only preset add/remove/update rewrite skills

preset enable, disable and set-priority update the registry without
rewriting skills, so they do not revert --no-model-invocation.

Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NRjt1UC5M2USX77TzRMemU

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Update openproject extension submitted by @ringind:\n- extensions/catalog.community.json (version, download_url, sha256, documentation)\n- docs/community/extensions.md community extensions table\n\nCloses github#4920\n\nAssisted-by: GitHub Copilot (model: gpt-5.2-codex, autonomous)\nCo-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Update adrkit extension submitted by @mbeacom:\n- extensions/catalog.community.json (version, download_url, sha256, compatibility)\n- docs/community/extensions.md community extensions table remains accurate\n\nCloses github#4926\n\nAssisted-by: GitHub Copilot (model: gpt-5.2-codex, autonomous)\nCo-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
…cific types (github#3917)

* fix: narrow exceptions in get_speckit_version() from Exception to specific types

Outer catch narrowed to PackageNotFoundError (the only expected
failure from importlib.metadata.version). Inner catch narrowed to
(OSError, ValueError, KeyError) for file/parse errors.

* fix: guard InvalidMetadataError and add regression tests

Addresses 2 Copilot findings:

1. Guard InvalidMetadataError (Medium): importlib.metadata.version()
   can raise InvalidMetadataError on malformed metadata. Mirrors the
   guard in _version._get_installed_version() — the exception class
   is looked up dynamically for Python versions where it is absent.

2. Add regression tests (Low): tests that FAIL against the previous
   bare except Exception implementation:

   - test_get_speckit_version_survives_invalid_metadata:
     InvalidMetadataError falls back correctly
   - test_get_speckit_version_reads_pyproject_fallback:
     PackageNotFoundError triggers pyproject.toml fallback
   - test_unrelated_exception_from_version_lookup_propagates:
     TypeError now propagates (was swallowed)
   - test_unrelated_exception_from_pyproject_propagates:
     RuntimeError now propagates (was swallowed)
   - test_pyproject_io_error_returns_unknown:
     OSError still caught (intended path preserved)

Assisted-by: GitHub Copilot (autonomous)
Update openproject preset submitted by @ringind:\n- presets/catalog.community.json (version, download_url, digest, documentation)\n- docs/community/presets.md community presets table\n\nCloses github#4921\n\nAssisted-by: GitHub Copilot (model: gpt-5.2-codex, autonomous)\nCo-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
…des (github#4923)

`and` and `or` evaluated both operands before combining them, so a guard
could not protect the operand after it. With an empty optional input,
`inputs.payload != '' and (inputs.payload | from_json | length)` still ran
`from_json` on the empty string and failed the run with "invalid JSON".

Return as soon as the left operand decides the result. The validation
probes behind condition corrections (`_evaluator_rejects`,
`_collect_leaves`) now pass a `_ProbeNamespace` as the outer namespace and
keep evaluating both operands, so an unknown name or misused filter in a
skipped operand is still reported.

Assisted-by: Codex CLI (model: gpt-6-astra, autonomous)
Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)
…ithub#4927)

* Update DocGuard — CDD Enforcement extension to v0.44.3

Update docguard extension submitted by @raccioly to the community catalog with the v0.44.3 release metadata and verified archive digest.

Closes github#4918

Assisted-by: GitHub Copilot (model: gpt-5.2-codex, autonomous)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Update DocGuard documentation URL

Co-authored-by: KSchlobohm <23503973+KSchlobohm@users.noreply.github.com>

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: KSchlobohm <23503973+KSchlobohm@users.noreply.github.com>
* feat: persist project symlink policy at command boundaries

Add strict per-invocation policy loading, init-only persistence, deny-mode configuration bootstrap, and explicit FileHelper handoff. Preserve init lifecycle rollback and config metadata compatibility with a bounded non-atomic overwrite operation. Other filesystem migrations remain deferred.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 661d79b5-8b35-41ce-b903-db64815502bc

* fix: validate init traversal before Windows presence probes

Reject missing and non-directory prefixes before parent traversal can be normalized into an existing init root. Preserve trusted existing directory and root aliases, propagate I/O failures, and cover the Windows behavior with portable regressions and native init cases.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 661d79b5-8b35-41ce-b903-db64815502bc

* test: cover artifact lookup policy-error boundaries

Verify missing and valid persisted policies permit lookup, while malformed settings and bootstrap links emit actionable JSON errors before installed-adapter or catalog access. Add seven net cases without changing runtime behavior.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 661d79b5-8b35-41ce-b903-db64815502bc

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 661d79b5-8b35-41ce-b903-db64815502bc
Update aee extension submitted by @tbitcs:\n- extensions/catalog.community.json (version, download_url, digest, metadata)\n- docs/community/extensions.md community extensions table\n\nCloses github#4935\n\nCo-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Update aee extension submitted by @tbitcs:

- extensions/catalog.community.json (version, download_url, sha256, and requirements)

- docs/community/extensions.md community extensions table (unchanged; entry already current)

Closes github#4937

Assisted-by: GitHub Copilot (model: GPT-5.2-Codex, autonomous)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
…ub#4942)

* fix(bundles): allow newer packaged patch releases with warnings

Permit stable patch upgrades of CLI-packaged components while retaining strict catalog pins and recording actual installed versions alongside requested pins.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 3dbacae3-752f-4608-a295-35d7c12fcee8

* fix(bundles): return structured compatibility warnings to CLI adapters

Address review feedback by collecting packaged patch warnings in InstallResult without domain-layer rendering, and presenting them only at bundle CLI boundaries.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 3dbacae3-752f-4608-a295-35d7c12fcee8

* test(bundles): cover packaged workflow validation and exact fallback

Add online and offline packaged workflow patch-warning coverage and prove that incompatible packaged releases require the exact pinned historical catalog release, rejecting absent or different releases.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 3dbacae3-752f-4608-a295-35d7c12fcee8

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 3dbacae3-752f-4608-a295-35d7c12fcee8
…4945)

Separate the common delivery and convergence loop from team-owned artifact maintenance choices. Clarify flow-back intent and preserve neutrality among persistence models.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 423e8bf6-30ec-49e3-bdb5-dc073f812ee5
…b#4931)

* fix(workflows): confine workflow and step catalog cache writes

Anchor catalog cache traversal and atomically replace JSON data and metadata without following repository-provided symlinks. Preserve discovery, freshness checks, refresh, and safe stale-cache fallback.

Add isolated clone/search regressions, race and failure coverage, and native Windows filesystem tests. Document the defensive cache behavior.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 85ee2fff-5185-406e-b4e5-c961e0ed0938

* fix(workflows): enforce Windows directory delete sharing

Include FILE_LIST_DIRECTORY in directory handle access so no-delete sharing participates in Windows share checks. FILE_READ_ATTRIBUTES alone allows empty cache directories to be renamed while held, as demonstrated by the three native Windows CI failures.

Keep the native Windows rename, ancestor pinning, and reparse-point regressions unchanged.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 85ee2fff-5185-406e-b4e5-c961e0ed0938

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 85ee2fff-5185-406e-b4e5-c961e0ed0938
Report the required title prefix and actual workflow run link before stopping, without changing matching-title validation or catalog policy.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 1f6b4b4d-ccdb-4f35-af4e-09c33d4a2ef2
* chore: bump version to 1.2.0

* chore: begin 1.2.1.dev0 development

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
* docs: distinguish extension reconciliation from core SDD

Clarify that core commands never reconcile feature artifacts while teams and extensions may provide reconciliation during or after delivery. Keep that behavior separate from post-delivery persistence.

Assisted-by: GitHub Copilot (model: GPT-5.6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 423e8bf6-30ec-49e3-bdb5-dc073f812ee5

* docs: preserve artifact-owning command roles

Clarify that core lacks an automatic all-artifact reconciler while contributors can rerun the command that owns each artifact. Use consistent during-delivery timing terminology.

Assisted-by: GitHub Copilot (model: GPT-5.6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 423e8bf6-30ec-49e3-bdb5-dc073f812ee5

* docs: retain converge remaining-work scope

Use the established remaining-work wording so converge coverage includes missing, partial, contradictory, and unrequested implementation outcomes.

Assisted-by: GitHub Copilot (model: GPT-5.6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 423e8bf6-30ec-49e3-bdb5-dc073f812ee5

* docs: qualify specify update targeting

Retain specify as an existing-spec update path while documenting that callers must target the existing SPECIFY_FEATURE_DIRECTORY to avoid creating a new feature directory.

Assisted-by: GitHub Copilot (model: GPT-5.6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 423e8bf6-30ec-49e3-bdb5-dc073f812ee5

* docs: keep reconciliation outside core commands

Remove artifact-command rerun prescriptions from persistence guidance. Leave reconciliation to documented team conventions or extensions while retaining analyze and converge boundaries.

Assisted-by: GitHub Copilot (model: GPT-5.6 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 423e8bf6-30ec-49e3-bdb5-dc073f812ee5

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 423e8bf6-30ec-49e3-bdb5-dc073f812ee5
* docs: add Korean (한국어) README

Add README.ko.md, a Korean translation structurally 1:1 with the English
README, link it from the language switchers in README.md, README.zh-CN.md
and README.ja.md, and add it to the markdownlint documentation allowlist.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: refine Korean README wording per review

- Translate needs-clarification / clarify as 명확화 (was 보완), matching
  the existing use of 명확화 for clarification in the SDD section.
- Rephrase the CLI setup sentence so it does not imply the terminal is
  used only for setup.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* ci: lint README.ja.md alongside the other localized READMEs

This PR modifies README.ja.md (language switcher), but the file was
outside the markdownlint DOC_GLOBS allowlist. It passes with 0 errors.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
kartsan03 and others added 27 commits November 9, 2026 12:01
* fix(presets): keep user files when removing preset skills

Preset skill cleanup removed the whole skill directory, so files the
user kept next to a generated SKILL.md (for example references/) were
deleted by `preset remove` and `integration switch`. Remove only the
generated SKILL.md and drop the directory when it is empty.

Assisted-by: Codex CLI (model: gpt-6.1-sol, autonomous)
Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)

* fix(presets): restore a removed skill directory's mode on rollback

Removing only SKILL.md and then the empty directory left the directory
removal out of the lifecycle journal, so a rollback recreated it with
default permissions. Journal the directory once it is empty, so its mode
comes back without snapshotting any user files.

Assisted-by: Codex CLI (model: gpt-6.1-sol, autonomous)
Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)

* fix(presets): don't follow a replaced skill directory during cleanup

Removing SKILL.md by path followed a skill directory that was replaced
by a link after validation, so the cleanup could delete a SKILL.md
outside the project. Delete it, and the empty directory, relative to
directory handles opened without following links below the project
(or global skills) root, and keep the journal calls on the same paths.
Without dir_fd, re-check each directory below the root before each
delete.

Assisted-by: Codex CLI (model: gpt-6.1-sol, autonomous)
Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)

* fix(presets): only skip expected errors in skill-file cleanup

unlink_in_directory() caught every OSError and ValueError, which also
hid journal failures (IntegrationInstallError is a ValueError) and real
delete errors such as PermissionError. Catch only the expected cases at
the operation that raises them and let everything else propagate, as it
did before this change.

Assisted-by: Codex CLI (model: gpt-6.1-sol, autonomous)
Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)

* test(presets): move helper and manager tests out of the command suite

The four tests that call unlink_in_directory() directly now live in
tests/specify_cli/integrations/test_file_changes.py, renamed after the
helper, and the global skills root test that calls PresetManager
methods moved to test_manager_skills.py. test_command_remove.py keeps
the tests that go through the CLI, as design/cli.md asks. No test
changed otherwise.

Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)

* fix(presets): recheck the skill directory after journaling it

The descriptor path compared the open skill directory with the entry
in its parent before journaling the removal, then removed the entry by
name. If the directory was replaced with another empty directory while
the journal ran, rmdir removed the replacement. Compare again right
after journaling, as the fallback path already does, and keep the
entry when it is no longer the directory that was emptied.

Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)

* test(presets): run the integration switch cleanup cases in the switch suite

The skill cleanup checks ran `preset remove` and `integration switch`
from the preset remove suite. Their bodies now live in
tests/specify_cli/presets/_skill_cleanup_helpers.py, and each command
suite runs them for its own command, as design/cli.md asks. No check
changed.

Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)
Assisted-by: Codex CLI (model: gpt-6.1-sol, autonomous)
- add ado v0.3.1 community bundle entry with validated metadata

- set sha256 from downloaded release asset

- add Spec Kit for Azure DevOps row to community bundles docs

Validation results:

- bundle ID and semver format: pass

- repository/doc/manifest checks: pass

- release URL pinning, HTTP 200, asset presence, and checksum verification: pass

- proposed catalog entry, component resolution, and checklist/testing evidence: pass

Closes github#5317

cc @praveenlavu

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Assisted-by: GitHub Copilot (model: gpt-5.3-codex, autonomous)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
* chore: bump version to 1.3.3

* chore: begin 1.3.4.dev0 development

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
…mmunity catalog (github#5348)

Add ado extension submitted by @praveenlavu to:
- extensions/catalog.community.json (alphabetical order)
- docs/community/extensions.md community extensions table

Closes github#5311

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…ithub#4876)

* feat(omp): support integration args and options in build_exec_args

- Pass per-step integration_args straight through to the omp CLI.
- Map the named options profile, thinking, and tools to their flags.
- Add validate_runtime_config to reject unknown options, empty values,
  and invalid thinking levels before the command runs.
- Reject integration_options.model so model selection stays on the
  command-step 'model' field.

* test(omp): cover integration args and options in build_exec_args

- Verify integration_args and integration_options are passed to omp
  and ordered after SPECKIT_INTEGRATION_OMP_EXTRA_ARGS.
- Accept every documented thinking level and valid runtime configs.
- Reject malformed args, the reserved model option, and invalid,
  unknown or non-string option keys and values.

* feat(omp): reject --model in integration_args and document OMP options

Model selection must go through the command-step `model` field, so a
`--model` flag in `integration_args` now fails validation before OMP is
launched. This matches the existing rejection of `integration_options.model`
and avoids two competing model sources on the command line.

- Reject `--model` and `--model=...` in `integration_args`; `--models`
  (Ctrl+P cycling scope) stays allowed.
- Document `profile`, `thinking`, and `tools` options, `integration_args`
  passthrough, and argument ordering in the integrations reference.
- Link the OMP options from the workflows per-step configuration section.
- Add tests for rejected `--model` forms and the allowed `--models` flag.

* fix(omp): reject '--' option terminator in integration_args

A bare '--' passed validation and, being inserted before the generated
--model, --mode json and prompt, made OMP treat them as positional
messages. Reject it in validate_runtime_config and add regression tests.

Assisted-by: Claude Code (model: Claude Sonnet 5.5, autonomous)
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* docs: drop OMP-specific paragraph from workflows reference

Integration-specific documentation lives only in docs/integrations/omp.md.

Assisted-by: Claude Code (model: Claude Sonnet 5.5, autonomous)
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix(omp): restrict integration_args to parser-safe tokens

OMP normalizes '--=' to the '--' terminator, and bare value-taking flags
(--system-prompt, --resume) consume generated flags or the prompt. Accept
only known valueless flags and self-contained --name=value tokens, and add
regression tests that replay OMP's parsing over the generated argv.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix(omp): reject equals-form help/version/print in integration_args

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix(omp): reserve --export/--alias/--mode and emit --mode text when JSON is off

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix(omp): reject inline values on valueless integration_args flags

OMP's parseArgs splits `--name=value` at `=`, matches only the flag
name for booleans, and drops the inline value, so `--yolo=false` and
`--auto-approve=false` still enable automatic approval. Reject the
equals-form of every known valueless flag; bare forms stay allowed.

Constraint: OMP parser (packages/coding-agent/src/cli/args.ts) ignores inline values on boolean flags
Rejected: Block only --yolo/--auto-approve | every boolean flag has the same misleading =false semantics
Confidence: high
Scope-risk: narrow
Assisted-by: Claude Code (model: Claude Opus 5.5, autonomous)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs(omp): clarify that integration `tools` is not an allowlist

- Reword the `tools` option description from "enable" to "select" to
  match how OMP actually treats the flag.
- Add a note that `tools` is not an exclusive allowlist or security
  boundary. OMP can still enable discovered custom and
  extension-registered tools that are absent from the list, so
  `tools: read` does not guarantee read-only access.
- Prevents users from relying on the option for sandboxing.

---------

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-authored-by: Ubuntu <khanh@dev-box.KHAS-MBOOK-1409A>
Co-authored-by: Kha Nguyen <khanh@truenas.local>
* Validate release tags before running release steps

Require the documented version format, pass versions through step environments, and cover the release sequence with local simulations.

Assisted-by: GitHub Copilot (model: gpt-6.1-sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Skip quoted loose-ref simulation on Windows

Limit the platform skip to the real-tag simulation and retain quoted-tag coverage in shell-only validation.

Assisted-by: GitHub Copilot (model: gpt-6.1-sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
* Use handle-based exclusive creation in FileHelper

Anchor byte and text creation to policy-resolved parent directories while preserving contained links, root aliases, modes, and encoding. Reuse the Windows handle primitive and cover creation behavior with focused tests.

Assisted-by: GitHub Copilot (model: gpt-6.1-sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Fix FileHelper creation handle compatibility

Restore the complete CreateFileW prototype, use search-only POSIX directory handles where available, and close previous parent descriptors during traversal. Add regression coverage for the Windows prototype, directory permissions, and bounded descriptor use.

Assisted-by: GitHub Copilot (model: gpt-6.1-sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Support directory search handles on older macOS Python builds

Use Darwin's native O_SEARCH value when Python does not expose the flag. Cover write/search-only parents with missing exported search flags and preserve no-follow directory traversal.

Assisted-by: GitHub Copilot (model: gpt-6.1-sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Clarify FileHelper creation access requirements and operation limits

Document Windows directory-listing and attribute access for pinned parents, and keep the path-based-operation disclaimer explicit for reads, directory creation, writes, overwrites, symlink creation, and deletion.

Assisted-by: GitHub Copilot (model: gpt-6.1-sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
…ption (github#5308)

* fix(extensions): show catalog entries missing name, version or description

`extension search` and `extension info` indexed a catalog entry's name,
version and description directly, so an entry without one of them made
both commands exit with a KeyError, and search never listed the entries
after it. Catalog entries are not validated field by field on purpose,
and `extension update` already skips a versionless entry (github#5108).

Read the three fields with .get(): `unknown` for a missing version (as
`extension add` and installed-extension info show it), the extension ID
for a missing name, and an empty description.

Assisted-by: Codex CLI (model: gpt-6.1-sol, autonomous)
Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)

* test(extensions): build the incomplete-entry project under tmp_path

The autouse fixtures request monkeypatch before temp_dir, so the cwd was
restored only after temp_dir's rmtree. Windows refuses to delete the
current directory (WinError 32), and every case errored at teardown.

Assisted-by: Claude Code (model: claude-opus-5-5, autonomous)
…railing newline (github#5357)

Co-authored-by: Claude <noreply@anthropic.com>
…ing bundles (github#5331)

* fix: preserve existing priority from registry when installing/refreshing bundles

When  or  is run, the
existing user-set priorities (via  /
) were being reset to the bundle's
declared priority. This fix reads the existing priority from the
registry first, and only falls back to the component's declared
priority or DEFAULT_PRIORITY if no existing priority is found.

Fixes: github#5318

* fix: compare plural component kinds and add priority regression tests

ComponentRef.kind is "presets"/"extensions", so the singular checks
never read the registry. Add tests covering registry-priority preservation
on install/refresh and fallback to declared/default priority.

* fix: normalize stored registry priorities before reuse

Registry data may hold strings, zero or booleans; normalize via
normalize_priority and fall back to the declared/default priority so
bundle refresh/update does not fail. Add regression tests for both
presets and extensions.

---------

Co-authored-by: Yi-111-a <yi-111-a@example.com>
… rollback version (github#5358)

packaging's Version.dev/.post are ints, so X.Y.Z.dev0 and X.Y.Z.post0 passed the stable-release check and the rollback hint pointed at vX.Y.Z.

Assisted-by: Claude Code (model: Claude Opus 5.5, autonomous)

Co-authored-by: goranai <28477490+goranai@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…e-new-feature.sh (github#5245)

* fix(scripts): treat option-like feature descriptions as text in create-new-feature.sh (github#5096)

* test(parity): fold full POSIX-quoted hints and cover apostrophe repo paths

Parse the complete POSIX shell word in fold_persistence_hints' bash branch
so shlex.quote's apostrophe spelling ('/tmp/O'"'"'Brien') canonicalizes to
the literal value instead of stopping at the first apostrophe and leaving
the escape suffix in the comparison. Apply the fold in
test_python_persist_hints_match_bash_for_spaced_repo_path as well, keeping
its explicit Bash-format assertion against the Bash output.

Fold before normalize_repo_paths so the canonicalized literal path can be
replaced across two-repo runs (normalize_repo_paths cannot match the split
Bash or doubled PowerShell apostrophe spellings). Add an apostrophe-path
regression case: a full-run parity parametrization, a persist-hints
end-to-end test, and a unit case pinning the seam parsing.

Regression evidence: before the helper fix, 4 tests failed (both apostrophe
full-run variants, the apostrophe persist-hints parity test, and the seam
unit case). After: targeted selection 8 passed;
tests/test_create_new_feature_python_parity.py 98 passed, 58 skipped
(PowerShell/Windows-only); all parity_helpers importers 438 passed,
97 skipped.

Assisted-by: Hermes Agent (model: mimo-v2.6-pro, autonomous)

---------

Co-authored-by: viwe-monai <viwe-monai@users.noreply.github.com>
…ithub#5332)

* Add T-Shirt Sizing Gate extension to community catalog

Add sizing extension submitted by @mauriziopinotti to:
- extensions/catalog.community.json (alphabetical order)
- docs/community/extensions.md community extensions table

Closes github#5304

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Fix sizing extension catalog metadata

Co-authored-by: mnriem <15701806+mnriem@users.noreply.github.com>

* Add sizing extension changelog URL

Co-authored-by: mnriem <15701806+mnriem@users.noreply.github.com>

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: mnriem <15701806+mnriem@users.noreply.github.com>
Co-authored-by: Ken Schlobohm <keschlob@microsoft.com>
* Update Spec Kit for Azure DevOps bundle to v0.3.2

Update the community catalog ado entry from 0.3.1 to 0.3.2.

Validation completed:
- Issue metadata and proposed catalog entry are consistent with bundle/bundle.yml
- Repository files present: bundle/bundle.yml, bundle/README.md, LICENSE
- Download URL pinning, ownership, and tag/version checks passed
- Release v0.3.2 contains ado-0.3.2.zip; archive download returned HTTP 200
- Submitted checksum matched downloaded archive
- Required catalog documentation and install/test evidence are present

Closes github#5344
cc @praveenlavu

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Assisted-by: GitHub Copilot (model: gpt-5.3-codex, autonomous)

* fix(bundles): refresh community catalog timestamp

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 520f069b-a04c-4ed8-be7c-c15bb5255811

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Ken Schlobohm <keschlob@microsoft.com>
Co-authored-by: Ken Schlobohm <23503973+KSchlobohm@users.noreply.github.com>
Copilot-Session: 520f069b-a04c-4ed8-be7c-c15bb5255811
…, Codex, and Claude Code (github#5249)

* feat(copilot): map workflow reasoning effort and context size

Translate the reasoning_effort and context_size workflow step settings
to --reasoning-effort and --context on both Copilot argv paths: prompt
steps through build_exec_args() and command steps through the
overridden dispatch_command(). The flags follow --model. Copilot CLI
validates the values itself.

Part of github#4916.

Assisted-by: GitHub Copilot (model: Claude Opus 5.5, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* feat(codex): map workflow reasoning effort and context size

Translate reasoning_effort to -c model_reasoning_effort="<value>" and
context_size to -c model_context_window=<n> after --model.
context_size must be a positive whole number no larger than the i64
range that Codex accepts.

Part of github#4916.

Assisted-by: GitHub Copilot (model: Claude Opus 5.5, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* feat(claude): map workflow reasoning effort to --effort

Claude Code only warns about an unknown --effort value and then runs
at its default effort, so Spec Kit checks the value against Claude's
documented levels (low, medium, high, xhigh, max) before dispatch.
The error names SPECKIT_INTEGRATION_CLAUDE_EXTRA_ARGS as an escape
hatch for newer levels. context_size is not mapped; its error points
to choosing a long-context model.

Part of github#4916.

Assisted-by: GitHub Copilot (model: Claude Opus 5.5, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* docs(workflows): list built-in reasoning effort and context size support

Add the support table for GitHub Copilot, Codex CLI, and Claude Code
to the workflow step reference and the shared CLI dispatch guide.

Part of github#4916.

Assisted-by: GitHub Copilot (model: Claude Opus 5.5, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix(integrations): omit rejected values from Claude and Codex errors

A dispatch setting can resolve from captured agent output, and a step
error is printed and written to log.jsonl, where captured text must
never appear. Claude Code's reasoning_effort check and Codex's
context_size check repeated the rejected value, so captured text that
passes the host's value pattern reached both. The errors now state the
setting and its rule without the value.

Part of github#4916.

Assisted-by: GitHub Copilot (model: Claude Opus 5.5, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix(copilot): validate dispatch setting values before dispatch

Copilot CLI rejects an unknown --reasoning-effort level or --context tier
with an error that repeats the value. A setting can resolve from captured
agent output, so a step that does not capture could print captured text
to the terminal. Spec Kit now accepts only the values that Copilot CLI
1.0.106 lists, and fails the step before the CLI starts. The error names
the accepted values and the SPECKIT_INTEGRATION_COPILOT_EXTRA_ARGS escape
hatch, but not the value.

Part of github#4916.

Assisted-by: GitHub Copilot (model: Claude Opus 5.5, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* docs(codex): keep captured output out of reasoning effort

Codex displays the reasoning effort it is given and accepts any value a
model advertises, so Spec Kit cannot check it against a fixed list. The
guide now advises setting it from a literal or a workflow input rather
than from captured output.

Part of github#4916.

Assisted-by: GitHub Copilot (model: Claude Opus 5.5, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* docs(workflows): run the context size example on Codex

The shared example sets context_size: "200000" on a step that inherits
the workflow's integration. Of the built-in integrations, only Codex
takes a token count, so the step failed when the workflow ran on GitHub
Copilot or Claude Code. The example now dispatches that step to Codex.

Part of github#4916.

Assisted-by: GitHub Copilot (model: Claude Opus 5.5, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* test(integrations): assert dispatch settings follow extra arguments

design/integration.md and docs/workflow-steps/types.md require the
dispatch-setting flags to follow the arguments from
SPECKIT_INTEGRATION_<KEY>_EXTRA_ARGS, but the tests set EXTRA_ARGS only
with both settings unset. The new tests set EXTRA_ARGS together with the
settings and check the full argv from build_exec_args() and
dispatch_command(): for Copilot in the skills and commands layouts, for
Codex, and for Claude Code, which maps reasoning_effort only.

Part of github#4916.

Assisted-by: GitHub Copilot (model: Claude Opus 5.5, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Sukanth Gunda <sukanthgunda@microsoft.com>
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
…ilters (github#5303)

* refactor(workflows): bound filter arguments and add a custom-filter registry

Replace FilterSpec.argument/bare with min_args/max_args and derive call-form
validation from one call_form_ok() shared by evaluation and static validation.
Built-in diagnostics keep their hand-written usage text byte for byte.

Add an immutable BUILTIN_FILTERS snapshot and register_custom_filter /
unregister_custom_filter(s) over the process-global FILTER_TABLE. Custom
handlers are called as handler(value, *args) and ordinary exceptions are
normalized to ValueError prefixed with the filter name. The expected-filters
sentence is built from the live table only when an error is raised.

Refs github#5285

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

* refactor(workflows): share package install, registry, URL and removal mechanics

Move the kind-independent parts of step package handling into reusable modules
so another installed package kind does not need a copy:

- _package_install: path/symlink safety, no-follow tree copy with excludes and
  budgets, the project lock, publish-with-rollback, removal, and one staged
  install_package flow, parameterized by InstallRequest, PackageSpec,
  PackageRules and StagingOps value objects.
- _package_registry: PackageRegistry (atomic, symlink-safe JSON registry with
  in-memory rollback); StepRegistry now only names its file and key.
- _archive_install: the --from URL flow (trust prompt, download, archive format
  agreement, safe extract, cleanup) and a shared error renderer.
- workflows/__init__: extract _import_package/_drop_modules from
  load_custom_steps.

The step installer keeps every established name as a thin binding and builds its
spec at call time, so existing patch points, messages and the standalone loading
used by the community step validator are unchanged. Existing step tests are not
modified.

Refs github#5285

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

* feat(workflows): load project-installed workflow functions

Add .specify/workflows/functions/<id>/ packages (function.yml + __init__.py).
The strict manifest parser, FunctionRegistry (a PackageRegistry subclass),
install rules (export names checked against built-ins, the registry and package
directories found on disk) and the loader are new; staging, copy, locking and
replacement come from the shared package install.

The loader clears the previous project's custom filters, validates every
manifest and reports all collisions before importing any code, imports each
package under a synthetic module name, and registers a package's exports
all-or-nothing. run, resume, workflow add and workflow update load functions
through one _load_project_workflow_plugins_or_exit helper that also renders
OSError. "functions" becomes a reserved workflow id.

Refs github#5285

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

* feat(workflows): add the specify workflow function commands

specify workflow function add (--dev or --from archive URL, --force), list
(registry metadata only, --json) and remove. add and remove use the shared
trust prompt, download checks, project lock and rollback; list never imports
package code. The three commands are registered in the MCP inventory as
currently unavailable.

Refs github#5285

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

* docs(workflows): document workflow functions

Add the workflow functions guide, the custom-filter section in the expressions
reference, the CLI command table and the design notes for the shared package
installation and the function loader.

Refs github#5285

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

* fix(workflows): cap custom filter argument counts at 16

A function manifest could declare an arbitrarily large max_args. It passed
install validation and was stored, and building the usage text then allocated
one string per declared argument, so function list and every workflow load
could exhaust memory.

Validate the bounds in one valid_arg_bounds() (0 <= min_args <= max_args <=
MAX_FILTER_ARGS) shared by the manifest parser, register_custom_filter and
function list. Oversized stored values are listed as 0..0 like other damaged
bounds.

Addresses review feedback on github#5303.

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

* fix(workflows): report non-string function.yml keys instead of raising

YAML mapping keys can be any scalar. Sorting and joining unknown manifest keys
assumed strings, so a numeric or mixed-type key raised TypeError, which
bypassed the loader's malformed-package handling and the CLI's
WorkflowFunctionError rendering.

Stringify unknown keys before sorting and reporting them, at the manifest and
function-item level.

Addresses review feedback on github#5303.

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

* test(workflows): make function tests portable to Windows and macOS

- Resolve the expected path in the remove-failure test; macOS and Windows temp
  directories are symlinked or 8.3 paths, so the unresolved path never matched
  what removal passes to rmtree and the injected error never fired.
- Skip the registry mode-bit assertion on Windows, like the existing
  mode-preservation test.
- Run the workflow-wiring shell fixtures through the current interpreter
  instead of POSIX printf, single quotes and ';' chaining, which cmd.exe does
  not interpret the same way.

Addresses review feedback on github#5303.

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

* test(workflows): clarify bare-condition validation test, cover wrapped loaded filter

The existing validation test asserted an error after the filter was loaded,
which is the bare-condition wrapping remediation, not a registry rejection;
rename it accordingly and add the wrapped counterpart that must validate clean.

Addresses review feedback on github#5303.

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

* fix(workflows): keep the zero-argument filter diagnostic in condition validation

_check_filter lost the "expected '| name' with no arguments or trailing
tokens" message when it moved to call_form_ok, so `from_json()` and `upper()`
were reported as a generic unsupported form. Restore the message and assert
it in full.

Refs github#5285

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

* fix(workflows): reject case variants of reserved function ids

On a case-insensitive filesystem `FUNCTION-REGISTRY.JSON` aliases the
registry file, so install published a package directory where the registry
must be written and then failed on save. Compare the id case-insensitively,
as validate_step_id does for step ids.

Refs github#5285

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

* docs,test(workflows): correct function docs and mirror tests to owning suites

Docs: state that filter names and argument counts in wrapped expressions
are checked at evaluation time, and which function-package load failures
abort the command versus skip the package.

Tests: move the shared package-registry suite to
tests/specify_cli/workflows/test_package_registry.py, the project-plugin
loading checks to test_plugin_loading.py, and the filter-table suite to
workflows/expressions/test_filters.py, per design/cli.md. The global
custom-function reset is now a shared opt-in fixture.

Refs github#5285

Assisted-by: pi coding agent (GitHub Copilot, model: claude-sonnet-5.5, autonomous)

---------

Co-authored-by: Markus <markus@example.com>
* Require HTTPS consistently for catalog sources

Align catalog source validation and documentation across all catalog types. Report preset configuration errors and use trusted HTTPS in installed-adapter tests.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Set explicit TLS minimum in catalog test server

Require TLS 1.2 or newer in the local HTTPS fixture and verify that the minimum is set explicitly rather than inherited from runtime defaults.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Align catalog redirect and fetch handling

Validate every integration catalog redirect, separate preset source configuration errors from recoverable fetch failures, and retain end-to-end loopback HTTP archive coverage with HTTPS catalogs.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Document catalog migration and cover preset search errors

Explain manual migration of legacy HTTP workflow and step catalog entries in project, user, and environment configuration. Cover fatal HTTP configuration and nonfatal fetch failures through the preset search CLI.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Strengthen catalog redirect tests and URL guidance

Fail preset search tests immediately if an HTTP redirect is unexpectedly accepted, and recommend HTTPS rather than HTTP in bundle catalog source guidance.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Clarify Python certificate trust for local catalogs

Distinguish Python TLS trust from browser and operating-system trust, document SSL_CERT_FILE CA-bundle configuration, and link local catalog guidance to the shared instructions.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Allow cleanup of legacy workflow catalog sources

Skip URL validation only for local workflow and step catalog removal while preserving priority ordering, configuration validation, and HTTPS requirements for discovery and fetching. Update migration guidance and cover legacy entries and invalid indexes through the CLI.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
* Use handle-based text overwrites in FileHelper

Anchor resolved overwrite destinations while retaining in-place updates, supported contained links, and native text behavior. Reuse shared Windows handle primitives and add overwrite regression coverage and contract documentation.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Preserve overwrite permission semantics across platforms

Honor the Windows creation mask, retain access through searchable unreadable POSIX directories, and inject overwrite failures at the shared descriptor boundary. Cover native creation-mask parity and directory permission success and failure cases.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Support search-only directory access on older Darwin builds

Use the Darwin kernel search bit when Python does not expose O_SEARCH or O_PATH, and exercise the missing-constant branch for existing and new overwrite destinations.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Match reference encoding in Windows creation-mask tests

Compare native creation with the overwrite API's UTF-8 default explicitly, including on Python versions whose Windows locale encoding differs.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Bound overwrite handles without changing the process mask

Let exclusive CRT creation apply Windows creation masks while retaining anchored existing-file opens. Close previous POSIX directory descriptors during traversal and clarify overwrite limitations. Cover interleaved calls, deep paths under low descriptor limits, and Windows creation collisions.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Keep native no-follow creation while honoring CRT permissions

Capture masked Windows creation attributes with an empty private probe in the pinned parent, then exclusively create the destination through a no-follow native handle. Clean up the probe, preserve existing in-place updates, and cover leaf collisions during attribute capture.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Avoid project delete-right requirements for Windows overwrites

Discover CRT-masked creation attributes in independent runtime-owned temporary storage instead of a disposable project sibling. Preserve native no-follow destination creation and cover create/write-without-delete directory ACLs.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Isolate overwrite ACL coverage from trusted-root setup

Apply create/write-without-delete permissions to the child destination directory, keeping root canonicalization separate from the overwrite permission regression.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Use exact delete permissions in the Windows ACL regression

Deny the advanced DE and DC rights explicitly and confirm ordinary file reading remains available before exercising overwrite behavior.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Document Windows overwrite directory access requirements

State that the canonical root and every resolved parent require directory-list and read-attributes access, including the resulting failure when an ACL allows file writes but denies listing.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Cover FIFO overwrites without a reader

Exercise leaf replacement with a no-reader FIFO in timeout-bounded subprocesses for both link policies. Require the nonblocking ENXIO failure and preserve the original file contents while retaining reader-present file-type coverage.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Cover overwrite capability and creation edge cases

Exercise missing and zero-valued primitives, regular files introduced before POSIX open, and Windows existing-file probe avoidance. Check explicit failures and preserve inode, hard-link, and permission behavior.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Simplify Windows overwrite concurrency coverage

Reuse the direct process-mask assertion while retaining concurrent calls and descriptor cleanup checks. Remove simulated mask state, synchronization events, and timed interleaving waits.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Update companion extension submitted by @alfredoperez:
- extensions/catalog.community.json (version, download_url, sha256, requires, provides, updated_at)

Closes github#5324

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Harden initial URL validation in HTTP helpers

Apply the shared HTTPS and loopback transport policy before resolving authentication or constructing requests. Cover accepted and rejected initial URLs and document the policy.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Classify URL policy rejections as definitive catalog failures

Share a URLPolicyError base between initial URL and redirect policy rejections. Preserve definitive errors in bundle catalog fetches and reference validation instead of treating them as temporary outages.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Clarify bundle URL policy failure documentation scope

Name online bundle catalog fetching and reference validation explicitly without implying changes to ordinary catalog lookup or cache policies.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Preserve URL policy errors across workflow catalog cache fallbacks

Bypass stale-cache fallback for shared URL policy rejections in workflow and step catalog fetchers while retaining the domain error contract and original cause. Cover expired cache, forced refresh, bundle reference rejection, and transport outage fallback.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Unify catalog URL rejection fallback behavior

Classify catalog callback and final URL validation rejections as URL policy failures so workflow and step catalogs never return stale data after rejected URLs, regardless of validator order. Preserve domain errors and ordinary fetch fallback, and extend expired-cache coverage.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Validate redirect headers before urllib normalization

Classify malformed Location and URI targets before urllib parses or normalizes them, covering every supported redirect status. Exercise the real redirect response entry point for expired-cache regressions, preserve valid relative redirects and auth handling, and explain the intentional URL-policy fallback change.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* Show clean URL policy reasons in catalog errors

Format URLPolicyError.reason in workflow and step catalog diagnostics while retaining the original exception as the cause. Assert exact messages across initial URL, redirect, and final URL rejections.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Keep Windows and macOS Python 3.11 coverage available through manual workflow dispatch.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
* fix: use handle-based FileHelper reads

Preserve supported contained links while opening reads through verified filesystem handles. Share Windows handle management, cover path replacement and descriptor cleanup, and document the read contract.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: refine FileHelper handle reads

Read directly from duplicated Windows handles, retain search-only directory access, and close superseded POSIX parents. Extend coverage for handle ownership, permissions, descriptor limits, and native Windows behavior.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* test: pass null handles to Windows fixture API

Use integer null arguments required by _winapi.CreateFile on Python 3.11.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: retain macOS search access with older Python SDKs

Use Darwin's native search-only flag when a Python build omits its binding, with permission regressions covering the missing-constant case.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* test: reuse native Windows reparse data in fixtures

Copy an OS-generated reparse buffer onto an empty source file and cover the fixture's native calls and cleanup paths.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: open Windows read children relative to validated handles

Preserve traversal-only directory access and strengthen permission-revocation, attribute-only junction, and ACL coverage while retaining the merged exclusive-create behavior.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* test: cover safe-read capabilities and clarify creation limits

Check each required POSIX read capability independently and document the Windows exclusive-create pathname limitation.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: reuse immutable Windows ABI types

Keep structure and pointer declarations at module scope so repeated child opens reuse type identities. Cover successful and failed opens plus native repeated-read lifetime checks.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* fix: initialize Windows ABI metadata lazily

Keep POSIX startup and filesystem operations independent of optional _ctypes while retaining one set of Windows ABI type definitions through the import cache.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

* refactor: share POSIX parent-directory traversal

Centralize directory capabilities, search flags, and descriptor ownership for reads, exclusive creation, and text overwrites. Preserve operation-specific leaf semantics and Windows paths.

Assisted-by: GitHub Copilot (model: GPT-6.1 Sol, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Resolves merge conflicts:
- pyproject.toml: merged MAS presets with upstream additions (bundles, constitution-sync, ruff lint)
- src/specify_cli/__init__.py: adopted upstream modular structure, added MAS imports
- src/specify_cli/command_init.py: ported MAS stack-aware init to upstream's refactored module

Version set to 1.4.0+mas.1 following MAS PEP 440 convention.

Co-authored-by: mbarrerarod <mbarrerarod@users.noreply.github.com>
- Add --stack and --skip-external-skills to init invocations
- Replace deprecated --no-git with --non-interactive/--skip-external-skills
- Skip tests that use --preset (mutually exclusive with --stack in MAS)
- 94 critical tests pass (20 MAS + 74 CLI tests)

Co-authored-by: mbarrerarod <mbarrerarod@users.noreply.github.com>
@coderabbitai

coderabbitai Bot commented Nov 12, 2026

Copy link
Copy Markdown

Important

  • Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

Run configuration

  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: Organization UI
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 7400e56a-6100-456b-9e11-131b2578b851
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

@mbarrerarod
mbarrerarod merged commit c2e2167 into main Nov 12, 2026
7 of 15 checks passed
if tmpl.get("name") == template_name and tmpl.get("type") == template_type:
file_path = tmpl.get("file")
if file_path:
manifest_candidate = pack_dir / file_path

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔒 Agentic Security Review

Severity: MEDIUM

Description: Preset template file paths are not contained to the preset directory on Windows. PresetManifest accepts a drive-relative path such as C: plus .. segments because it only rejects os.path.isabs or a normpath that starts with .., and _manifest_declared_template then joins that string onto the preset directory with no anchor or relative_to check. On Windows, ntpath.join either appends those .. segments under the preset directory or, when the drive differs, drops the preset directory and resolves against that drive's cwd. Install then reads the outside file and copies its bytes into agent skill files, and into agent command files when the command uses append/prepend/wrap or when reconciliation treats the escaped path as a non-preset source.

Impact: A preset author is untrusted package input, but install runs as the user and is supposed to stay inside the preset directory. On Windows, one manifest entry can read a chosen file the user can read (for example an SSH key or cloud credential) and write those bytes into agent command or skill files that the coding agent loads. Extension manifests already reject this shape of path; preset templates do not, so installing a local, archive, or catalog preset crosses the package-root boundary.

Remediation: Validate preset template file paths with relative_extension_path_violation (reject anchors, backslashes, and .. under both POSIX and Windows semantics). Before any read or agent registration, require the candidate to stay under the preset directory with candidate.resolve().relative_to(pack_dir.resolve()), and do not register a layer whose resolved path fails that check.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.