By name: verity-prism-pr-description description: Use when creating a pull request for a Verity Prism implementation phase branch — reads the phase spec and audit doc to produce a complete, accurate PR description rather than guessing context from memory.
Verity Prism — Phase PR Description
Overview
Generates PR descriptions for feat/phase-N-* branches by reading primary sources: the approved spec, the audit findings, and the live commit list. Avoids guessing finding descriptions or test counts.
Process
Run these in order before writing a single word of the PR body:
- Read the spec:
docs/superpowers/specs/<date>-phase<N>-*-design.md— pull the exact findings in scope, decisions made, what was deferred. - Read the audit doc:
docs/code-audit-2026-05-29.md— pull the finding severity, one-line description, and the exact "Fix" instruction for each finding being closed. - Get commit list:
git log main..HEAD --oneline - Get test delta:
docker-compose run --rm -e TEST_DATABASE_URL=... -e SECRET_KEY=... backend pytest tests/ 2>&1 | tail -1— note the before/after count (check prior PR for the "before" number). - Check for migrations:
git diff main...HEAD --name-only | grep alembic/versions— if any, note what DDL they install and thatalembic upgrade headis required on deploy.
PR Description Template
## Summary
- **<FINDING-ID>** — <one-line from audit doc, finding severity in parens>. <what was done, 1 sentence>.
[repeat for each finding]
## Migration notes (if any)
Migration `<revision_id>` — <what DDL it installs>. Requires `alembic upgrade head` on deploy before starting the app.
## Test delta
<N_before> → <N_after> tests (+<delta> new). All passing.
New test files: `<list any created test files>`
## Findings closed
| Finding | Severity | Description |
|---|---|---|
| <ID> | <Critical/High/Medium/Low> | <exact description from audit doc> |
[repeat]
## Test plan
- [ ] Full suite: `docker-compose run --rm -e TEST_DATABASE_URL=postgresql://catalyst:catalyst@db:5432/catalyst_test -e SECRET_KEY=<strong-key> backend pytest tests/ -v` → <N> passed
- [ ] <one specific verification per finding, e.g. "UPDATE audit_log raises InternalError">
- [ ] Ruff lint clean: `ruff check backend/app/ backend/tests/`
🤖 Generated with [Claude Code](https://claude.ai/claude-code)
Quick reference
| What to include | Where to get it |
|---|---|
| Finding IDs + severity | docs/code-audit-2026-05-29.md |
| Decisions / deferred scope | Phase spec file |
| Commit list | git log main..HEAD --oneline |
| Test count | Run the suite; check prior PR for baseline |
| Migration revision ID | git diff main...HEAD --name-only | grep alembic |
| Deferred items | Spec "out of scope" section |
Common mistakes
Writing from memory — Finding descriptions drift from the audit doc's exact language. Always copy from the source.
Omitting migration impact — Any phase that adds an Alembic revision requires an alembic upgrade head note. The app will error on next deploy if skipped.
Wrong test count — Use the actual suite output. The baseline count is in the previous phase's PR, not in your head.
Missing deferred scope — The spec lists what was intentionally left out. Including this in the PR prevents reviewers from filing "why didn't you also fix X?" comments.