SuperAPI AI Directory

retrospective

Agents & models

View repository

Updated 10-23 23:08

About this skill

Your objective is to run the periodic closed-loop quality retrospective.

Commands

  • $retrospective status: read .cocoplus/lifecycle/retrospective-ledger.jsonl and report last run, pending requests, and open actuator proposals.
  • $retrospective run: analyze recent session artifacts, classify failure modes, propose the cheapest sufficient actuator, and append the result to the ledger.

Failure Taxonomy

Use exactly one tag per attributable failure:

  • present-not-consulted
  • present-contradicted
  • absent-via-truncation
  • absent-via-never-retrieved
  • absent-via-compaction
  • cant-tell

Actuator Hierarchy

Prefer the cheapest sufficient actuator:

  1. hook or gate,
  2. skill or command,
  3. structural agent-level rule,
  4. standing context.

Do not propose a standing-context addition unless the cheaper layers cannot enforce the rule.

Exit Criteria

  • Token-weighted wasted effort is reported.
  • Every applied edit has a verbatim after anchor for future verification.
  • The ledger records evidence, classification, proposed actuator, and decision.

Anti-Rationalization

Shortcut / Temptation Why It Fails
Treat the skill as complete because the file exists Skill contracts must describe observable behavior and verification, not just command names.
Skip artifact and safety checks for a small command Small commands still mutate state or guide execution; preserve the same gates.