| name | talos-drift-sweep |
| description | Use when checking Talos for cross-repo drift, stale `.agent` propagation, duplicated protocol logic, broken submodule assumptions, or inconsistent docs and code boundaries. Prefer existing verification scripts and boundary checks, summarize concrete drift findings, and separate confirmed issues from unverified risk. |
Talos Drift Sweep
Use this skill for anti-drift and repository health checks.
Load:
references/checks.md
../planning/program-anchor-index.md
Workflow:
- Identify the likely drift class: schema duplication, cross-repo import leak,
stale
.agent sync, stale docs, or submodule mismatch.
- Use the existing checks first before inventing a new one.
- When a check fails, trace the exact file, module, or generated artifact that
drifted.
- Separate confirmed breakage from likely but unverified risk in the summary.
- If the fix touches multiple repos or submodules, name the required follow-up
explicitly.
Guardrails:
- Do not describe “drift” abstractly without naming the concrete duplicated or
stale artifact.
- Do not assume all submodules are meant to move together; verify ownership and
intended boundaries.
- Do not create new generated content flows when an existing sync script already
owns the artifact.
Done checklist:
- Drift class identified.
- Existing checks run or consciously skipped with reason.
- Findings are concrete and scoped.
- Follow-up repo or submodule actions called out.