Skip to content

feat(planning): name the architecture skills in design and plan - #16422

Merged
kyle-sexton merged 3 commits into
mainfrom
feat/planning-architecture-skills-and-later-line
Nov 15, 2026
Merged

kyle-sexton merged 3 commits into
mainfrom
feat/planning-architecture-skills-and-later-line

Conversation

@kyle-sexton

@kyle-sexton kyle-sexton commented Nov 15, 2026 •

Copy link
Copy Markdown
Member

No related issue: owner decision from a design interview on how planning uses the architecture skills (2026-11-14), recorded outside this repository.

Summary

No planning skill named /architecture:principles, and only /planning:brainstorm named /architecture:improve, so a design that changed module structure used neither the module-design vocabulary nor the friction scan.

  • /planning:design invokes /architecture:principles (if enabled) when the work adds, splits, merges or moves a module or its interface, and names /architecture:improve (if enabled) for modules that already exist. Without the architecture plugin it says once what the plugin would add.
  • /planning:plan names both skills in its prerequisite check when the plan reshapes existing modules.

Fix

  • plugins/planning/skills/design/SKILL.md: one paragraph in Phase 4 and two rows in the relationship table.
  • plugins/planning/skills/plan/SKILL.md: one clause on the "Which other skills apply?" line.
  • Two outcome cases under plugins/planning/evals/.

Verification

  • claude plugin eval --scaffold, plugin arm only, 3 runs per case: design-module-split-names-architecture-skills 3 of 3, plan-reshape-modules-names-architecture-skills 3 of 3. No baseline arm was run: each prompt is a /planning: invocation.
  • The plan case can fail: run against the base version of the plugin (the same case, without this PR's line in plan/SKILL.md) it scored 0 of 3. Its one grader matches the text architecture:improve, so it tests that the line is followed, not how well.
  • The eval session holds the planning plugin alone, so the design case exercises the not-enabled branch (the plugin is named, not invoked). The enabled branch has no case.
  • plugins/planning/tests/interview-defenses.test.sh: PASS=185 FAIL=0; plugins/planning/scripts/repin-digests.test.sh: 23 passed.

Scope change

The first head also added a Later: line to the /planning:interview round header. That paragraph sits in a section whose digest interview-defenses.test.sh pins, so Test Bash failed. It is removed from this PR and ships separately with its re-pin.

Related

  • #16089 added /architecture:principles.
  • #16418 also adds a fragment under .changes/planning/; the two touch different files.

🤖 Generated with Claude Code

…a Later line

/planning:design invokes /architecture:principles when the work changes
module structure and names /architecture:improve for existing modules;
/planning:plan names both in its prerequisite check. A
/planning:interview round header gains one Later: line for the topics
of rounds already known to follow.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@kyle-sexton
kyle-sexton marked this pull request as ready for review November 15, 2026 19:43
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 15, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-15T19:44:39.472105Z 4abf59f Draft marked ready
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

The paragraph sat in the interview skill's Stance section, whose digest
interview-defenses.test.sh pins. This change no longer edits the
interview skill; the Later line needs a re-pin and ships separately.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@kyle-sexton kyle-sexton changed the title feat(planning): name the architecture skills in design and plan, add a Later line feat(planning): name the architecture skills in design and plan Nov 15, 2026
@github-actions

Copy link
Copy Markdown
Contributor

Lane run report

Field Value
Activity pr-refine/fix-docs
Model rung 0
Leg handoff
Run job success
Activity step success
Commits pushed none (read lane)
Run 43754010253

Failed GitHub API calls

Method Path Status
GET /repos/{repo}/git/ref/tags/* 404

@kyle-sexton kyle-sexton added the do-not-merge Hard merge gate: do not merge while applied. label Nov 15, 2026
@github-actions

github-actions Bot commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

claude-security-review reviewed 8c9c8e8223035753c5cf18e55b3acce429a7e444: read 2 of 2 changed file(s); 0 finding(s), 0 as inline comments.

Security review: PR #16422, head 8c9c8e8 (incremental since a33fec0)

I found no security issues in this push. This lane already reviewed an earlier head, so this review was blocking only: it reports only CRITICAL and IMPORTANT findings.

Scope: I read every hunk of both files in .claude-lane/incremental.diff.

What changed: the push adds two eval files to the design-module-split-names-architecture-skills case.

  • graders/no-code-edit.md is a tool_used grader. It fails a run that calls Edit with an input matching \.py", and it applies to both arms.
  • samples/no-code-edit.json holds the grader's pass and fail fixtures. The pass fixtures are an edit to a .work/.../design-threads.md file and an empty answer. The fail fixture is an edit to billing/invoicing.py.

What I checked:

  • Executable content: neither file holds a script, a command or a network call, and neither touches a token or secret. Both are static grader configuration and fixture data.
  • Instruction surface: no standing instruction is deleted, narrowed or softened.
  • Trust controls: neither file matches a trusted-path glob, and no workflow, action, merge-hold script or plugin manifest changed.

Spend (not a finding): the added grader runs inside the existing eval case, so it adds no model runs beyond the prepaid usage that case already draws.

critical-findings: 0

@github-actions github-actions Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

claude-review: 1 inline finding(s) on a33fec0d550d9e5c455a58556b082bd72558dc25; the summary is in this lane's pull request comment.

claude-lane: claude-review

@github-actions

github-actions Bot commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

claude-review reviewed 8c9c8e8223035753c5cf18e55b3acce429a7e444: read 2 of 2 changed file(s); 0 finding(s), 0 as inline comments.

Code review (incremental): PR #16422 at 8c9c8e82

Scope: this push only. I read every hunk of both files on the incremental list. This lane has already reviewed an earlier head, so this pass reports blocking findings only.

What changed

  • plugins/planning/evals/design-module-split-names-architecture-skills/graders/no-code-edit.md is new. It is a tool_used grader on Edit with input_match: '\.py"', min: 0, max: 0 and arm: both.
  • plugins/planning/evals/design-module-split-names-architecture-skills/samples/no-code-edit.json is new. Its two pass samples are an Edit to .work/split-invoicing/design/design-threads.md and an empty answer. Its fail sample is an Edit to billing/invoicing.py.

Checked and fine

  • The push fixes this lane's earlier finding: the case had no Edit grader, so its "No .py file is written or edited" outcome was only half checked. That thread is resolved and settled.
  • The new grader follows the case's own no-code-write.md and the sibling design cases, for example design-fact-placed-once/graders/no-code-edit.md.
  • Each sample sorts correctly under \.py". The .md edit does not match, so it passes. The billing/invoicing.py edit matches the .py" closing quote in the serialized input, so it fails.
  • The author says validate-cases.py reports no FAIL. That is unverified: I had no shell to run it.

No new findings.

Verifier dropped 0 of 0 candidates.

@kyle-sexton kyle-sexton removed the do-not-merge Hard merge gate: do not merge while applied. label Nov 15, 2026
…ites

The case's expected outcome says no .py file is written or edited, and
only Write was graded. A no-code-edit grader matches the sibling design
cases.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@kyle-sexton
kyle-sexton added this pull request to the merge queue Nov 15, 2026
Merged via the queue into main with commit 738abf8 Nov 15, 2026
50 checks passed
@kyle-sexton
kyle-sexton deleted the feat/planning-architecture-skills-and-later-line branch November 15, 2026 20:05
@github-actions github-actions Bot added the merge-path: owner-agent Post-merge accounting record (README.md, Merge paths) label Nov 15, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

merge-path: owner-agent Post-merge accounting record (README.md, Merge paths)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant