fix-next-test

Instructions

This is a thin resume wrapper around fix-tests, not an independent workflow. It processes one durable root-cause group, then leaves the campaign ready for the next invocation.

Directus is the canonical test state and claim store. Use scripts/tests.py commands for current failures, triage, claims, and reruns. Do not use test-results/*.json as the source of truth; those files are import/export artifacts only.

Step 1: Create or resume campaign state

python3 scripts/tests.py campaign start --session <session-id> --json
python3 scripts/tests.py campaign status --campaign <campaign-id> --json
python3 scripts/tests.py campaign next --campaign <campaign-id> --lease --session <session-id> --json

If a specific spec was supplied, pass its canonical key with campaign start --test-key playwright::<spec-name> when creating a new campaign. Do not bypass campaign persistence.

• <spec-name>: Override auto-pick — fix this specific spec instead
• --skip-session: Skip sessions.py start (already in a session)
• --rerun-only: Just rerun all currently-failed tests and report results, don't fix anything

Step 2: Execute one group through the shared contract

Follow fix-tests Steps 2-4 exactly:

python3 scripts/tests.py campaign prepare --group <group-id> --expected-behavior "..." --criterion "..."
python3 scripts/tests.py campaign attempt --group <group-id> --approach "..." --outcome <outcome>
python3 scripts/tests.py run --campaign <campaign-id> --group <group-id>
python3 scripts/tests.py campaign complete-group --group <group-id> --commit <sha>

Persist acceptance criteria before source edits and every attempt after it is known. Verify all group members, not only the first test returned by triage.

If a scoped sessions.py deploy integrates the fix before the group rerun, the current worktree is closed for subject-commit-bound evidence. Start a fresh sessions.py session/worktree, then rerun the same campaign/group command from that new session. Do not retry campaign runs, Docker restarts, gate-deploy checks, or proof-video commands from a worktree after the routing guard reports it is already merged.

Step 3: Return campaign state

Print campaign status after the group becomes green or blocked. Do not claim the campaign is complete unless its status is completed; newly exposed child groups remain part of the same campaign.

Rules

• One root cause per invocation — process one durable group, then return campaign state
• Never run vitest/playwright locally — always dispatch via scripts/tests.py run
• Directus campaigns are canonical — do not maintain separate chat or local progress state.
• Attempts are durable — inspect prior attempts and do not repeat rejected approaches.
• Read before writing — always read the failure report and source code before changing anything
• Console errors are real bugs — fix them in app code, never suppress
• Keep live-probe failures visible — group them under their provider parent incident and continue independent groups
• Block structurally — required unresolved input keeps the campaign active and blocked, never partially successful.