Skip to content

chore: declare a model and effort for each agent skill stage - #110

Merged
V3RON merged 5 commits into
mainfrom
claude/elegant-hamilton-l2yerq
Sep 24, 2026
Merged

V3RON merged 5 commits into
mainfrom
claude/elegant-hamilton-l2yerq

Conversation

@V3RON

@V3RON V3RON commented Sep 24, 2026 •

Copy link
Copy Markdown
Contributor

Closes #109

What changed

Each agent stage now runs on its declared model and effort, whether work-issue delegates it, a person types it as a slash command, or another agent loads it. Review, triage, design and memory stages fork on opus at high effort. implement-issue forks on sonnet at medium, and e2e-device on sonnet at low. file-issue (sonnet, medium) and cut-release (sonnet, low) stay inline because they need the conversation. work-issue now invokes each stage's skill directly through the Skill tool, with the task as arguments. It no longer starts a general-purpose agent to load the skill, and it never passes a model. The Skills table in AGENTS.md gains Model, Effort and Forked columns, and the table says to change them together with the frontmatter. The mapping follows the issue's amended table, including the effort addendum.

Acceptance criteria

# Criterion Test Tier
1 Every skill's frontmatter matches the table (model, effort, fork) Checked against the amended table manual
2 AGENTS.md Skills table has Model, Effort and Forked columns matching the frontmatter Checked against the frontmatter manual
3 CI check for drift Dropped at the maintainer's request: drift can't break anything silently —
4 work-issue invokes each stage with the Skill tool and arguments and still gets the report block Probes below: each fork returned its skill's report-block first line to the caller integration
5 Headless runs prove the routes Probes below integration

Headless probes, Claude Code 2.1.281, claude -p … --output-format json in this branch. Bash, Edit, Write and Agent were disallowed. Each probe asked the forked skill to echo a nonce from its arguments and then stop.

Session Route modelUsage / effort Fork's reply
--model sonnet Skill tool → review-pr claude-sonnet-5 (caller) + claude-opus-5-5 (fork) N7Q2 claude-opus-5-5 PR: #N Verdict: approve | comment | request-changes
--model opus Skill tool → implement-issue claude-opus-5-5 (caller) + claude-sonnet-5 (fork) K4M9, Sonnet 5 (claude-sonnet-5), Branch: issue-N-slug PR: #M (draft)
--model sonnet --effort high Skill tool → e2e-device, PreToolUse hook logging effort caller's Skill call: high; fork's Read call (agent_type: general-purpose): low PROBE E5T1 + Target(s): <list> Commit: <sha>

The nonces (N7Q2, K4M9, E5T1) show that each forked skill received the arguments it was invoked with.

E2E evidence

Not applicable. This changes agent tooling only.

Checklist

  • CHANGELOG.md has an entry under Unreleased (writing-changelog skill), or the change is not user-visible
  • User-facing docs updated for every surface the change touches (writing-user-docs skill), or the change is not user-visible
  • No new import past a module's index.ts; no new direct node:* I/O outside an adapter
  • Simplification checklist from the architecture skill applied, exceptions explained above
  • docs/ARCHITECTURE.md updated if a surface it describes changed

Out of scope

  • The branch is claude/elegant-hamilton-l2yerq, not issue-109-<slug>, because this session had to push to that branch.
  • The review and E2E stages were skipped at the requester's direction. This PR has had no agent review.

Status

Implement: done (4/4, criterion 3 dropped) Review: skipped (by request) E2E: not applicable Ready: yes

🤖 Generated with Claude Code

https://claude.ai/code/session_01WwCFb6cH1WSzCS8iob6A7e

Red: no skill declares a model and the table has no Model or Forked
column yet. Refs #109.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WwCFb6cH1WSzCS8iob6A7e
Review, triage, design and memory stages fork on opus; implement and
E2E fork on sonnet; file-issue and cut-release declare sonnet inline.
work-issue invokes each stage through the Skill tool with arguments
instead of a general-purpose agent. AGENTS.md gains Model and Forked
columns. Closes #109.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WwCFb6cH1WSzCS8iob6A7e
Red: no skill declares effort and the table has no Effort column yet.
Refs #109.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WwCFb6cH1WSzCS8iob6A7e
Review, triage, design and memory run at high; implement and file-issue
at medium; E2E and cut-release at low. AGENTS.md gains an Effort column.
Closes #109.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WwCFb6cH1WSzCS8iob6A7e
The frontmatter and the AGENTS.md table change together by hand; a
drift between them cannot break anything silently.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WwCFb6cH1WSzCS8iob6A7e
@V3RON
V3RON merged commit 438f5e5 into main Sep 24, 2026
4 checks passed
@V3RON
V3RON deleted the claude/elegant-hamilton-l2yerq branch September 24, 2026 15:10
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.

Declare a model for each agent skill stage

2 participants