Menü
Cosecha aprendizajes, valida evidencia y produce un cierre de conversacion reproducible con handoff, riesgos y plan de actualizaciones durables.
| name | cierre-conversacion |
| version | 0.2.0 |
| description | Cosecha aprendizajes, valida evidencia y produce un cierre de conversacion reproducible con handoff, riesgos y plan de actualizaciones durables. |
| owner | JM Labs |
| triggers | ["cierre-conversacion","session-audit","cosechar-aprendizajes"] |
| allowed-tools | ["Read","Grep","Glob","Bash"] |
Use this skill to close a long or explicitly ended conversation without losing useful state. It turns the session into a deterministic closeout packet: what was decided, what changed, what remains open, which learnings are reusable, which validations are real, and which durable updates are only proposed unless authority is explicit.
Activate when any of these conditions is true:
cierre-conversacion, session-audit, closeout, retrospective, handoff, or learning harvest.Do not activate for generic filesystem cleanup, unrelated summarization, or requests to erase history without preserving required evidence.
pass when closure evidence is enough, block when it is not.Classify the trigger as explicit, threshold-based, or manual audit. If activation is ambiguous, produce a short proposal rather than claiming a completed closeout.
Collect only observable facts from the conversation, files, commands, PRs, or user-provided context. Tag every claim as [CÓDIGO], [CONFIG], [DOC], [INFERENCIA], [SUPUESTO], or [POR_CONFIRMAR].
Fill the required sections in this order:
Do not update tasklog, changelog, memory, or skill assets unless the current authority is explicit. When authority is missing, return the proposed update text and mark it [POR_CONFIRMAR].
Before declaring pass, ensure no completed task relies only on [SUPUESTO] or [POR_CONFIRMAR], every validation claim has command or artifact evidence, and no failed validation is hidden behind a pass decision.
Use assets/activation-policy.json, assets/output-contract.json, assets/evidence-policy.json, assets/harvest-checklist.json, and assets/durable-update-policy.json as the deterministic source of truth for activation, output shape, evidence tags, harvest checks, and durable update authority.
[POR_CONFIRMAR] and keep Guardian blocked.Deterministic checks live in scripts/. Run:
bash skills/cierre-conversacion/scripts/check.sh
The script validates closeout report fixtures offline and requires invalid fixtures to fail.
session-end-cleanuptasklog-managementchangelog-managementpre-compact-context[CÓDIGO]: local files, commands, PR metadata, CI output, or diff evidence.[CONFIG]: user instructions, repository policy, or workflow contract.[DOC]: stable project documentation.[INFERENCIA]: derived conclusion from evidence.[SUPUESTO]: unverified assumption.[POR_CONFIRMAR]: pending user, CI, remote, or external confirmation.durable_update_plan.Redacta agradecimientos post-proceso diferenciados por persona, evidencia de interaccion, voz de marca y lint offline contra FOMO, hustle, servilismo y promesas no verificables.
Optimiza CV y carta de presentacion para ATS, rol objetivo y voz de marca con lint offline de keywords, secciones, longitud, contacto y tono.
Orchestrates JM-ADK Alfa repository maintenance in fixed safety phases: bootstrap, repo sync audit, local state preservation, isolated branch planning, selective import or consolidation planning, cleanup planning, validation gates, and closeout blockers.
Validates plugin directory layout against the official Claude Code plugin spec. Triggers: validate structure, check layout, directory audit, plugin skeleton check.
Validates frontmatter and body structure of all skills, agents, and commands in a plugin. Triggers: validate components, check frontmatter, component audit, skill validation.
Use when the user asks to "scan for secrets", "detect credentials", "sanitize sensitive data", "check for exposed passwords", "run security gate G0", or mentions secret detection, credential scanning, security gate G0, sensitive data masking, API key exposure, token detection.