| name | loop-harness-verification |
| description | General-purpose harness for scheduled repo work — stage in an isolated worktree, verify with a genuinely separate second session, ship only after a pass. Read-only — reports, never delivers. |
loop-harness-verification procedure
The prompt carries intent; this skill carries the procedure — keep it a checklist,
not an essay. Adapted from Forward Future Loop Library #17 (Istasha) to this
library's L1-first, report-only shape.
- Read — the wrapped task's inputs (CI status, issue backlog, dependency
manifest, docs — whatever this instance covers) and
loop-harness-verification-state.md's last run.
- Stage — in an isolated worktree, produce a candidate patch or outbox
message for the wrapped task.
- Verify, genuinely separately — invoke a second, distinct session (the
loop-verifier agent, or a second claude -p / opencode run call —
never the same pass that staged the output) to check the candidate
against explicit criteria.
- Decide:
- Pass → write
harness-report.md recommending delivery. Do NOT
deliver it yourself.
- Fail → preserve the findings in
loop-harness-verification-state.md
and retry, but only within the configured retry limit.
- Write
harness-report.md: source revision, staged output, verifier
result, delivery status, next scheduled run — the source's own required
fields.
- Record — update
loop-harness-verification-state.md, append one
line to loop-run-log.md, commit both.
- Take NO other action — never deliver, merge, or apply anything. This
loop is L1 report-only until a human promotes it.