SuperAPI AI Directory

loop-run-record

General & other

View repository

Updated 10-16 05:50

About this skill

Loop Run Record

Record what happened in one externally executed run. This skill is a post-run evidence boundary, not an executor or a new readiness judge.

Installed dependencies

The recorder reuses loop-readiness-score, which in turn requires goal-engineering and methodology-selector. Confirm that all four skills are installed under the same skills root. If any sibling is missing, stop and report the complete installation set.

Preconditions

Read references/run-record-contract.md.

  • Goal input, rendered goal, readiness input, and readiness result must be the exact artifacts that produced a ready verdict.
  • The observation must contain one result for every validation check.
  • Actual evidence, hypothesis outcome, feedback, next action, and a human-return reason must be explicit.

Stop on stale, forged, incomplete, blocked, or supervised artifacts. Do not repair them.

Workflow

  1. Reassess the supplied goal with the supplied readiness envelope.
  2. Require the fresh assessment to match the supplied result and identity.
  3. Match each observed check to one canonical validation check.
  4. Derive yes, partial, or no from check statuses; do not trust a caller-supplied overall result.
  5. Bind expected evidence, actual evidence, hypothesis outcome, feedback, next action, and human-return reason into one deterministic record.
  6. Return the record with review-required.
node /absolute/path/to/loop-run-record/scripts/create-run-record.mjs /absolute/path/to/input.json

Resolve the first path from the directory containing this SKILL.md. The agent prepares the JSON input; do not ask the human to hand-write it.

Boundaries

  • Do not execute, retry, resume, schedule, or monitor a task.
  • Do not modify the goal, readiness result, methodology, skill, or evidence.
  • Do not claim that a record authorizes commit, merge, deploy, publication, or another run.
  • Do not create memory or update a playbook automatically.
  • Do not accept blocked or supervised readiness as evidence that a run was permitted.

Stop

Stop after one canonical record or one explicit validation failure. Return the record to a human for the next decision.