docs: AI Hero steering-section vetting — lanes 7–9 decisions, invocation-mode rubric, harness-claims verdicts - #2990
Merged
Conversation
…n-by-section decomposition
Steering-section session of the AI Hero course effort. The v1.2 lane-5
rejection ('at parity or stronger') held only for the pruning/audit half;
the authoring half carries three gaps now tracked as course lanes 7-9
(#2909 authoring-time doctrine, #2910 invocation-mode rubric, #2911
steering-surface validations). Adds the decomposition table to the SSOT,
corrects v12-map row 25, records drift finding 6, and lands the steering
lane index as a topic addendum.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbfCrj3X9FfGL7VRZYmrn4
…n contract + decision rows Interview rounds 1-2 closed (register gate clean, user confirmed 2026-08-17): new model-invoked write-side skill homed in docs-hygiene, scope generalized to any agent-consumed markdown, trigger reliability gated by a shipped eval suite (hook rejected as overengineering), completion criteria adopted at both moments, two-loads doctrine in the skill body with a philosophy cross-reference. STEERING.md carries the lane 7 decision rows for lane 6's harvest. Refs #2909. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QbfCrj3X9FfGL7VRZYmrn4
…uild issues filed, SSOT annotated Research run (dispatched, artifact + coverage gates exit 0, fresh-context verifier graded, project fit applied per the contract's claim ladder) produced the 25-surface Claude Code enumeration + 13 ecosystem analogues, committed as the durable design doc. Build filed as #2962 (skill + eval suite + philosophy cross-ref) and #2963 (skill-quality completion-criteria criterion). SSOT decomposition annotated design-locked; leading-words tracked strand carries its retirement disposition. Closes #2909. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QbfCrj3X9FfGL7VRZYmrn4
…ric homed, fleet re-grade decided, enforcement filed Interview locked (8/8, register gate clean, brief=ok): model-invoked default with three exception classes (side-effect/manual-timing, setup, maintainer-only), homed as the convention-registry owner doc docs/conventions/invocation-mode/README.md with PLUGIN-PHILOSOPHY cross-refs (setup contract, Instruction economy). Router pattern REJECTED with reason (composition-router carve-out kept); upstream's user-invoked default REJECTED (#693, cloud scope, multi-repo discoverability). ADR 0005-bounded re-grade executed in-lane over the 10 non-setup true skills: 9 KEEP, 1 FLIP. Enforcement filed as #2968 (explicit key on 17 skills + skill-quality criterion + skill-authoring pointer), flip as #2969. SSOT gap-3 rows dispositioned; invocation-reach strand CONFIRMED against current docs, release trigger retired, audit trigger kept. Closes #2910. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QbfCrj3X9FfGL7VRZYmrn4
…tch, eval fixture adopted Five user-confirmed decisions from the lane 9 (#2911) interview round 1: STEERING.md-rows-only artifact shape, C5 carve-out + C7 note criteria patch, C3 nested-CLAUDE.md row + @-mention distinction, design-smell caveat homed in remediation guidance, /init-then-prune eval fixture adopted against the audit eval suite. Harness-claims verdicts and parity rows follow research-gated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QbfCrj3X9FfGL7VRZYmrn4
…s-tool discovery convention User-raised, verified directionally: Codex CLI, Cursor, Gemini CLI, and VS Code Copilot share the .agents/skills project + user discovery paths; the spec defines only the folder format. Claude Code keeps its own .claude paths; whether it also reads .agents stays with #2911's harness-claims bundle. Refs #2909. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QbfCrj3X9FfGL7VRZYmrn4
… patch and eval fixture filed Round-2 decisions user-confirmed (all recommendations accepted): 7-claim harness verdict table recorded against Claude Code v2.1.233 with verifier-corrected corroboration labels (claim 4 single-pool docs-only with cure note; ~/.agents/skills equivalence refuted two-pool and marked do-not-repeat); auto-memory lesson recorded PARITY with no work; L1-2 leftovers recorded as verdict-backed rows only; "highway / stale highway" handed to lane 6. Implementation filed: criteria patch (C5 carve-out, C7 stale-highway note, C3 nested-CLAUDE.md + @-mention rows, design-smell remediation) and the /init-then-prune static eval fixture. Steering lanes 7-9 all closed; chain close (PR + merge) is next. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QbfCrj3X9FfGL7VRZYmrn4
|
Bugbot is not enabled for your account, so this pull request was not reviewed. Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 6838c14a21
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".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
…bric per review Contract-slice prune gate compliance (topic-docs convention, graduation step 3): git mv the chain's four contract-tier files out of docs/topics/ into their durable homes — the lane 7 Brief and verified surface enumeration plus the lane 8 Brief into docs/specs/, and the steering decision record into docs/upstream/aihero-steering-lanes.md (lane 6 harvests it into aihero-course.md and dissolves it). All repo references updated to the new paths. Review fixes on docs/conventions/invocation-mode/README.md (both verified before applying): qualify the skillOverrides name-only lever by origin — it does not reach plugin skills, so the marketplace levers are description trimming and /plugin enablement; recount the fleet at the chain-close merge — 215 top-level skills = 141 false / 17 missing key / 57 true (the four skills merged in from main are explicit false and default-conformant; the 10-skill grade stands). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QbfCrj3X9FfGL7VRZYmrn4
kyle-sexton
pushed a commit
that referenced
this pull request
Aug 17, 2026
Main's #2990 confirmed the invariant against official docs and retired the upstream-release trigger; this PR's #2940 pass recorded the fleet audit and rewording. The merged strand keeps both records, with the audit-side re-trigger as the only remaining trigger per the lane 8 disposition. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CnzwTKoTa6xNY7iyEzMYpm
|
Claude finished @kyle-sexton's task in 31s —— View job
|
10 tasks
|
Claude has reviewed this PR 1 time. The lane skips further automatic reviews after 5; deleting this comment resets the count. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No linked issue
Summary
Records the AI Hero course steering-section vetting: three interview-first lanes (authoring doctrine, invocation-mode doctrine, steering-surface validations) discussed, decided, and closed, with every decision graduated to durable-tier docs and implementation deferred to filed work items per the chain contract (lanes decide, they do not implement).
Fix
Docs-only. The lane work, graduated to durable paths per the topic-docs convention (contract slices under
docs/topics/are branch-only; pre-prune revision6838c14acarries them at their original paths):docs/upstream/mattpocock-skills.md+mattpocock-skills-v12-map.md— superseded the blanketwriting-for-agentsparity claim with a section-by-section decomposition; SSOT strands dispositioned (leading-words strand routed to the build; invocation-reach invariant CONFIRMED against current docs).docs/upstream/aihero-steering-lanes.md— lane index and full decision rows for lanes 7–9, including the lane-9 harness-claims verdict table (7 course claims graded against Claude Code v2.1.233 with corroboration labels;~/.agents/skillsequivalence refuted and marked do-not-repeat). Lane 6 (docs(upstream): aihero course lane 6: shared vocabulary adoption and aihero-course provenance doc #2904) harvests this intoaihero-course.mdand dissolves it.docs/specs/write-for-agents-brief.md+docs/specs/agent-doc-surfaces.md— lane 7's locked design contract and the verified 25-surface agent-doc enumeration (contract for docs-hygiene: implement write-for-agents — authoring-time skill for agent-consumed markdown #2962/skill-quality: add completion-criteria check criterion + pointer to write-for-agents doctrine #2963).docs/specs/invocation-mode-doctrine-brief.md— lane 8's locked Brief (contract for skill-quality/fleet: enforce the invocation-mode rubric — explicit disable-model-invocation on 17 skills, check criterion, skill-authoring pointer #2968/planning: flip questionnaire to model-invoked — the one invocation-mode re-grade flip #2969).docs/conventions/invocation-mode/README.md— the adopted invocation-mode rubric (model-invoked default + three exception classes, inverting upstream's user-invoked default with recorded evidence).Review fixes (both Codex findings verified against the repo before applying): the
skillOverrides: "name-only"lever is now origin-qualified (it does not reach plugin skills — the marketplace levers are description trimming and/pluginenablement), and the fleet grade totals recounted at the chain-close merge (215 top-level skills = 141false/ 17 missing key / 57true; the four skills merged in frommainare explicitfalse, so the 10-skill grade stands).Implementation rides filed issues: #2962/#2963 (lane 7 build + audit criterion), #2968/#2969 (lane 8 enforcement +
planning:questionnaireflip), #2987/#2989 (lane 9 criteria patch + eval fixture).Verification
scripts/check-contract-slice-prune.sh --check-diff origin/main: passed — no path left underdocs/topics/.npx markdownlint-cli2over all changed docs files: 0 issues.plugins/planning/scripts/check-open-questions.sh; lane 9 register in bullet-row form, all questions answered).origin/mainmerged into the branch pre-PR; diff against main is docs-only (no plugin files touched).Related
Refs #2909, #2910, #2911 (closed lane issues this PR makes durable); Refs #2962, #2963, #2968, #2969, #2987, #2989 (filed implementation work items); Refs #2904 (lane 6 harvests the steering decision record); Refs #1419 (contract-slice graduation discipline this PR follows).
🤖 Generated with Claude Code
https://claude.ai/code/session_01QbfCrj3X9FfGL7VRZYmrn4