Repository navigation
Add composer skill and selection fixtures - #31
Merged
Merged
Conversation
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
This was referenced Oct 20, 2026
Include composer in the missing-skills doctor expectation. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
Clarify non-English probe limits, draft export gates and skill handoffs. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
Continue unverified banks through available reviews and report specific draft blockers before delivery. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds composer for turning a brief into a behaviour-led course outline and handing writing and review to the owning skills. Includes five synthetic brief/selection fixtures, a hard-gate selection scorer, CLI regressions and reviewed Claude/Codex adapter hashes.
Feedback contrasts and any teaching added to support them must be supported by the brief's sources, including each condition. Fidelity checks those teaching pages directly against the sources and returns unsupported teaching or wrong conditions to the writer. Banks where masked scoring does not apply remain unverified, but continue the plain and cover probes, key check, explanations probe, fidelity review, alignment and accessibility checks, and report. The person is asked at the end, including for unattended runs. Single-family hosts still use a fresh-context reviewer, preferring a different model. One label, "Draft", lists the specific blockers in the outline and report. Every draft gets a temporary HTML export solely for praxity-check; results are reported and the export is deleted without shipping. Deliverable exports wait until all blockers clear.
All seven Opus P3s addressed: French-only register questions, precise learners-guide location, delayed export handoff, explicit dontKnow=false when unselected, narrower skill routing, README skill list and removal of the nonexistent nightly-check claim. HTTPS source links remain allowed. No P3 fixes deferred.
Checks passed:
node --test test/composer.test.mjs test/skills.test.mjs test/manifest.test.mjs(112 tests, including adapter hashes and relative links)node test/composer/score-selection.mjs --self-testnpm run validategit diff --checkNew guidance regressions failed before these revisions. The earlier fix also corrected the doctor expectation and rejected dangling barrier references, invalid record uses, and blank or misdirected environment fixes. No full suites or model-driven course generation run locally; CI handles integrated checks.
Installed prax-format still claims briefs until Studio #1814 merges and Rubato re-pins Studio; the owner will sequence that follow-up.
No nightly workflow yet. It needs API keys and an owner-approved budget.
Review: 2/4/8; fixes 2.
No further review rounds unless a P1 recurs. Draft PR; no merge requested.