Test All Fixer

Run the repository validation loop from the repo root. Fix the actual problem reported by the first failing check, then rerun the full suite until every test passes or an external blocker stops progress.

Workflow

1. Locate the repository root by finding tests/test_all.sh.
2. Run ./tests/test_all.sh from that root. If the user says tests/tests_all.sh, treat that as the real tests/test_all.sh path in this repo.
3. Read the first failing Python script in tests/ and the file named in the traceback or error message.
4. Make the smallest change that satisfies the lint's intent. Do not weaken or disable tests unless the user explicitly asks for that.
5. Rerun ./tests/test_all.sh after each fix. Keep iterating until it prints All tests have passed. or you hit a blocker that requires user input.
6. Report the failing check you fixed, the files changed, and whether the full suite is now clean.

Guardrails

• Prefer full-suite reruns over single-test shortcuts. It is fine to run an individual failing script once while debugging, but always finish with ./tests/test_all.sh.
• Preserve user changes. Avoid broad cleanup outside the failing check's scope.
• When the failure comes from lint_comp_template.py, treat tests/diff_base/ as the baseline, but confirm whether template_ws is the intended source of truth before changing workspace files.
• Respect repo rules in AGENTS.md, especially docker/compose.yaml naming, required default files, and README/docs parity.
• Use IGNORED_WORKSPACES only when the user asks to exclude workspaces or when a temporary workspace would otherwise make the suite meaningless.
• If a fix to template_ws implies syncing many workspaces, keep the immediate test fix minimal unless the failure already proves broader drift that must be corrected.
• Do not use {PLACEHOLDER_MULTILINE} unless the user explicitly asks for it. Prefer explicit placeholder lines such as {PLACEHOLDER_#} when updating tests/diff_base/.

Failure Routing

Use references/test-map.md to map each failing test script to the files and invariants it checks.

Output

State the exact command run, the failing script that was addressed, the files modified, and whether the final full-suite run passed.