TGPSKI · TGPSKI · Aug 9, 2026 · Aug 9, 2026
@@ -0,0 +1,12 @@
+---
+name: Agent work item
+about: Minimal issue body for agent-consumable units of work
+title: ""
+labels: agent-work
+---
+
+## Context
+- SKILL.md
+
+## Work
+- [ ]
@@ -10,16 +10,23 @@ format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 - `PATTERN.md`: the Directed Contexts pattern specification — vocabulary,
   invariants, context kinds, router contract, ownership matching, route
   cases, and instruction-file classification.
-- `SKILL.md`: the `generate-directed-contexts` skill router with operating
+- `SKILL.md`: the `directed-contexts` skill router with operating
   modes for bootstrap, adopt, recover, audit, and decline.
 - `references/`: phase modules for coordinate check, boundary model,
   generation, validation, adoption of existing instruction files, and
   lifecycle maintenance.
 - `scripts/contextctl.go`: stdlib-only Go tool with `scan` (deterministic
   JSON inventory), `check` (Markdown contract validation), `routes`
   (route-case evaluation), and `drift` (codebase/context-set divergence).
+  `scan` reports `instruction_surfaces.primary_surface` — the root file an
+  agent reads first, resolved by precedence: Context Router, root `SKILL.md`,
+  root `AGENTS.md`, root harness alias.
 - `assets/templates/`: skeletons for the root router, context index, domain
-  context, policy context, and the target-repo lifecycle skill.
+  context, policy context, the target-repo lifecycle skill, and the agent
+  work-item issue template.
+- `.github/ISSUE_TEMPLATE/agent-work-item.md`, and Phase 3 Step 7 generating
+  the same template for target repos with the Context line resolved from the
+  scan.
 - Golden examples for
   [abductive-triage](https://github.com/TGPSKI/abductive-triage) (semantic
   ownership) and

@@ -1,7 +1,5 @@
 # Contributing
 
-Thanks for your interest in Directed Contexts.
-
 ## Ground rules
 
 - **The pattern is the product.** `PATTERN.md` is canonical. Changes to the
@@ -44,6 +42,7 @@ A new golden example needs:
 
 ## Reporting issues
 
-Use GitHub issues. For validation bugs, include the offending Markdown and
-the `contextctl` output. For boundary-modeling discussion, a route case that
-routes wrongly is worth a thousand words.
+Use GitHub issues; `agent-work-item` is the template for agent-consumable
+work. For validation bugs, include the offending Markdown and the `contextctl`
+output. For boundary-modeling discussion, include a route case that routes
+wrongly.
@@ -84,7 +84,7 @@ candidates.
 | [SKILL.md](SKILL.md) | The generator skill router |
 | [references/](references/) | Phase modules: coordinate check, boundary model, generate, validate, adoption, lifecycle |
 | [scripts/contextctl.go](scripts/contextctl.go) | Stdlib-only Go tool: `scan`, `check`, `routes`, `drift` |
-| [assets/templates/](assets/templates/) | Skeletons for the router, domain/policy contexts, index, and lifecycle skill |
+| [assets/templates/](assets/templates/) | Skeletons for the router, domain/policy contexts, index, lifecycle skill, and issue template |
 | [examples/abductive-triage/](examples/abductive-triage/) | Golden example: semantic ownership in an all-Markdown repo |
 | [examples/security-context-spec/](examples/security-context-spec/) | Golden example: mixed spec/tooling/workflow ownership |
 

@@ -1,5 +1,5 @@
 ---
-name: generate-directed-contexts
+name: directed-contexts
 description: "Generate bounded codebase ownership contexts that agents can load inline or spawn in isolation. Analyzes a repository, proposes ownership boundaries, and produces a root AGENTS.md router plus .subagents/ context modules with executable validation. Use when asked to create directed contexts, split a large AGENTS.md, set up subagent contexts, add an agent routing table, or audit an existing context set."
 license: GPL-3.0
 compatibility: "Requires Go toolchain (1.22+) for the contextctl scan/check/routes/drift tool."
@@ -69,7 +69,7 @@ Inspect the scan output's `instruction_surfaces` to select an operating mode.
 | Router and contexts exist                           | **Audit** — run `contextctl check` + `drift`, then `references/lifecycle.md` |
 | Native `.claude/agents/` or `.github/agents/` exist | Preserve them; classify personas vs. ownership contexts per `PATTERN.md`, then continue detected mode |
 | Nested `AGENTS.md` files exist                      | Decide retain-vs-consolidate per `references/adoption-existing-agents.md`, then continue |
-| Repository too small for meaningful partitioning    | **Decline** — recommend one concise `AGENTS.md`; do not manufacture contexts |
+| Repository too small for meaningful partitioning    | **Decline** — recommend one concise `AGENTS.md`; do not manufacture contexts. The Phase 3 Step 7 issue template still applies |
 
 **Too small** means any two of: fewer than ~30 source files, one package root,
 one entry point, one audience. Say so plainly and stop.
@@ -91,8 +91,8 @@ one entry point, one audience. Say so plainly and stop.
 | ----- | ---------------------------------------- | -------------------------------------------------- | ------------------------ |
 | 1     | `references/phase-01-coordinate-check.md` | Verify repo coordinates before modeling anything   | Nothing                  |
 | 2     | `references/phase-02-boundary-model.md`   | Propose ownership boundaries with weighted evidence | Proposal table only      |
-| 3     | `references/phase-03-generate.md`         | Generate contexts, index, router, lifecycle skill  | The context set          |
-| 4     | `references/phase-04-validate.md`         | Mechanical checks and route-case validation        | Route-case fixtures      |
+| 3     | `references/phase-03-generate.md`         | Generate contexts, index, router, lifecycle skill, route cases, issue template | The context set |
+| 4     | `references/phase-04-validate.md`         | Mechanical checks and route-case validation        | Nothing                  |
 | —     | `references/adoption-existing-agents.md`  | Rules for existing AGENTS.md / nested files / personas | —                    |
 | —     | `references/lifecycle.md`                 | Audit, drift, split/merge maintenance loop         | —                        |
 
@@ -108,7 +108,7 @@ Each phase inherits from its predecessors — never re-ask for these:
 | --------------------------- | -------------- |
 | Repo root and package roots | Phase 1        |
 | Source vs. generated/vendor classification | Phase 1 |
-| Instruction-surface inventory | Phase 1      |
+| Instruction-surface inventory, including the primary surface | Phase 1 |
 | Confirmed boundary table    | Phase 2        |
 | Context names and owned paths | Phase 2      |
 | Generated file list         | Phase 3        |
@@ -122,6 +122,7 @@ Each phase inherits from its predecessors — never re-ask for these:
 | `assets/templates/domain-context.md`         | Primary domain context skeleton       |
 | `assets/templates/policy-context.md`         | Policy overlay skeleton               |
 | `assets/templates/maintain-directed-contexts/SKILL.md` | Lifecycle skill for the target repo |
+| `assets/templates/agent-work-item.md`        | Target-repo issue template            |
 
 After generation, always run:
 

@@ -1,6 +1,6 @@
 # {REPO_NAME} — Agent Instructions
 
-<!-- Context Router. Generated by generate-directed-contexts; maintained by humans.
+<!-- Context Router. Generated by the directed-contexts skill; maintained by humans.
      This routing table is canonical — tooling parses this Markdown directly. -->
 
 {ONE_PARAGRAPH_REPO_DESCRIPTION}

@@ -0,0 +1,12 @@
+---
+name: Agent work item
+about: Minimal issue body for agent-consumable units of work
+title: ""
+labels: agent-work
+---
+
+## Context
+- {PRIMARY_INSTRUCTION_SURFACE}
+
+## Work
+- [ ]
@@ -24,7 +24,11 @@ skill, not by any domain.
 
 - `source.json` — pinned source commit metadata
 - `inventory.json` — deterministic `contextctl scan` snapshot at that commit
-- `expected/` — the complete generated context set
+- `expected/` — the complete generated context set, issue template included
+
+The issue template's Context line is `AGENTS.md`, not the source repo's
+pre-generation primary surface (`SKILL.md`): the router this run generates
+outranks it. See `references/phase-03-generate.md`, Step 7.
 
 ## Validate
 

@@ -0,0 +1,12 @@
+---
+name: Agent work item
+about: Minimal issue body for agent-consumable units of work
+title: ""
+labels: agent-work
+---
+
+## Context
+- AGENTS.md
+
+## Work
+- [ ]
@@ -7,6 +7,7 @@
   "languages": [],
   "entry_points": [],
   "instruction_surfaces": {
+    "primary_surface": "SKILL.md",
     "root_agents_md": true,
     "has_routing_table": false,
     "nested_agents_md": [],

@@ -26,7 +26,7 @@ keep new files inside automatically owned.
 
 - `source.json` — pinned source commit metadata
 - `inventory.json` — deterministic `contextctl scan` snapshot at that commit
-- `expected/` — the complete generated context set
+- `expected/` — the complete generated context set, issue template included
 
 ## Validate
 

@@ -0,0 +1,12 @@
+---
+name: Agent work item
+about: Minimal issue body for agent-consumable units of work
+title: ""
+labels: agent-work
+---
+
+## Context
+- AGENTS.md
+
+## Work
+- [ ]
@@ -25,6 +25,7 @@
     "tools/validate"
   ],
   "instruction_surfaces": {
+    "primary_surface": "AGENTS.md",
     "root_agents_md": true,
     "has_routing_table": false,
     "nested_agents_md": [],

@@ -1,7 +1,7 @@
 ---
 name: adoption-existing-agents
 description: "Rules for adopting a repository that already has instruction files: monolithic AGENTS.md, nested AGENTS.md hierarchies, CLAUDE.md aliases, native personas, and skills."
-parent: generate-directed-contexts
+parent: directed-contexts
 metadata:
   author: TGPSKI
   version: "0.1"

@@ -1,7 +1,7 @@
 ---
 name: lifecycle
 description: "Audit and maintenance loop for an existing context set: drift detection, split/merge decisions, review-date hygiene."
-parent: generate-directed-contexts
+parent: directed-contexts
 metadata:
   author: TGPSKI
   version: "0.1"
@@ -40,6 +40,7 @@ go run scripts/contextctl.go drift --repo /path/to/target/repo
 | Stale review date                | Re-read the guide against the code; fix or confirm, then re-date |
 | Verification command missing     | Replace with a command that exists; never leave aspirational commands |
 | New entry point / trust surface  | Check whether an overlay's activation table covers it         |
+| Issue-template Context line points at a moved or deleted file | Repoint it at the current primary instruction surface |
 | Context above split threshold    | Step 3                                                        |
 | Context below merge threshold    | Step 4                                                        |
 

@@ -1,7 +1,7 @@
 ---
 name: phase-01-coordinate-check
 description: "Verify the repository's coordinate system — root, package layout, source classification, instruction surfaces — before any boundary modeling."
-parent: generate-directed-contexts
+parent: directed-contexts
 metadata:
   author: TGPSKI
   version: "0.1"
@@ -32,7 +32,9 @@ The scan establishes, without asking the user:
   `pyproject.toml`, `Cargo.toml`, `pom.xml`, `Makefile`, …);
 - languages and manifests;
 - source versus generated/vendor/build trees;
-- existing `AGENTS.md`, `CLAUDE.md`, `CODEX.md`, skills, and native agents;
+- existing `AGENTS.md`, `CLAUDE.md`, `CODEX.md`, skills, and native agents,
+  and which of them is the repository's primary instruction surface
+  (`instruction_surfaces.primary_surface`);
 - CODEOWNERS and ownership documentation;
 - entry points, binaries, APIs, and deployment units;
 - build, test, lint, and validation commands (from Makefile targets and
@@ -95,6 +97,7 @@ A short coordinate summary the user confirms, carried forward to Phase 2:
 - Package roots: {list}
 - Source trees: {list} | Generated: {list} | Vendored: {list}
 - Instruction surfaces: {list, with router/context/skill/persona classification}
+- Primary instruction surface: {path, or "none"}
 - Declared ownership: {CODEOWNERS summary or "none"}
 - Build/test commands: {list}
 - Unresolved questions: {none, or the questions asked above}

@@ -1,7 +1,7 @@
 ---
 name: phase-02-boundary-model
 description: "Construct candidate ownership domains from weighted evidence, resolve ambiguous boundaries with competing hypotheses, and produce a confirmed boundary table."
-parent: generate-directed-contexts
+parent: directed-contexts
 metadata:
   author: TGPSKI
   version: "0.1"
@@ -91,6 +91,18 @@ Rules the table must satisfy:
 - Every source path from Phase 1 is either owned or deliberately left
   unowned with a stated reason.
 
+Present the file set the table implies alongside it, so the user sees every
+path Phase 3 will write before confirming:
+
+| File                                          | What it is                              |
+| --------------------------------------------- | --------------------------------------- |
+| `AGENTS.md`                                   | Router; replaces or absorbs any existing root file |
+| `.subagents/README.md`                        | Context index                           |
+| `.subagents/AGENTS-{DOMAIN}.md`               | One per row of the table above          |
+| `.subagents/route-cases.json`                 | Route fixtures                          |
+| `.agents/skills/maintain-directed-contexts/SKILL.md` | Lifecycle skill, if accepted     |
+| `.github/ISSUE_TEMPLATE/agent-work-item.md`   | Issue template, unless the repo is not on GitHub |
+
 ## Step 5: Confirm
 
 **Decide**: Present the table. Ask the user **only** about medium- or