メニュー
Run the full local gate (lint, tests+coverage, eval regression check, README/baseline consistency, clean tree) before pushing — use before opening a PR or when the user asks "is this ready to push?".
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
SOC 職業分類に基づく
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/Juanpacol/SEPHIROTH --skill release-checkコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
SKILL.md を表示中
Run the RAG evaluation harness and report metric deltas — use after touching data/rag, the Evidence Agent, or Citation Guard, or whenever the user asks to check/update the eval numbers.
Add a new clinical guideline document to the RAG corpus with matching golden/paraphrase eval cases and tests — use when the user wants to expand data/rag/SEED_GUIDELINES with a new topic.
Launch SEPHIROTH locally and smoke-test the core flow (health, auth, one consultation) — this repo's project verify skill, with this machine's port/venv quirks baked in.
| name | release-check |
| description | Run the full local gate (lint, tests+coverage, eval regression check, README/baseline consistency, clean tree) before pushing — use before opening a PR or when the user asks "is this ready to push?". |
Run each step in order and stop at the first failure — report exactly which check failed and why, don't silently continue past a red step.
cd clinical-ai-copilot
.venv/bin/ruff check .
.venv/bin/ruff format --check .
If ruff format --check fails, that's auto-fixable: .venv/bin/ruff format . then re-review the diff before continuing (formatting changes should be whitespace/quote-only — if anything looks semantically different, stop and investigate).
PYTHONPATH=.:platform .venv/bin/pytest --cov
Must show Required test coverage of 87.0% reached and all tests passing. A coverage drop below 87% blocks — don't lower the threshold in pyproject.toml to make it pass; add tests for the uncovered lines instead (--cov-report=term-missing will show exactly which lines).
PYTHONPATH=.:platform .venv/bin/python -m intelligence.evaluation.run --mode ci
Must print Overall: PASS with no staleness warning. If it warns the baseline is stale, data/rag/SEED_GUIDELINES or intelligence/evaluation/datasets/golden.json changed without a matching --mode full --record run — use the /eval skill to refresh before continuing.
Compare the metric table in README.md's ## Evaluation section against intelligence/evaluation/results/latest.json. If any number differs, the README wasn't updated after the last eval refresh — fix it now (this is a quick, common miss).
.venv/bin/bandit -c pyproject.toml -r . -ll -ii
Must report zero HIGH-severity + HIGH-confidence issues (this mirrors CI's security job exactly). MEDIUM/LOW findings are advisory — mention them but don't block on them.
.venv/bin/pip-audit -r requirements.txt
Non-blocking — report anything with an available fix, but don't hold up the push on CVEs with no fix yet.
git status --short
Everything intended for this change should be staged/committed; nothing stray (stray __pycache__, .next/, or other build artifacts should already be gitignored — if you see any tracked, that's a regression in .gitignore coverage worth flagging).
Green on steps 1-5 (6 is advisory) means the branch is safe to push and open a PR. Summarize which steps passed; if anything failed, name the exact command and output that failed rather than a general "something's wrong."