titan-retirement-playbooklisted

# Titan Retirement Playbook Retiring a titan is a first-class lifecycle event. Done poorly it wastes the institutional knowledge the titan accumulated; done well it makes the corpus stronger for successors. This playbook defines the canonical procedure. ## 1. When titan classes deprecate Deprecation is triggered when **any one** of these four signals fires: ### (a) Mission no longer recurs The titan's mandate was episodic — a one-off v0.8.3 integration, a bounded quality push, a marketing sprint. Once the deliverable is in production and no further dispatch is planned, the titan is a dormant class occupying PROMPT-BRAIN real estate. Signal: zero dispatches in the last two major releases with no roadmap entry. ### (b) Subsumed by another titan or a generic skill A second titan was built that covers a strict superset of this one's mandate (e.g., `playwright` subsuming a hypothetical `puppeteer` titan), or the work proved simple enough to be encoded as a reusable skill (`.claude/skills/`). Signal: a newer titan or skill explicitly cites this one as its predecessor in its own SKILL.md. ### (c) Consistently underperforms Across two or more runs the titan scores <70/100 in READY-REPORT, ships deferred CRITICALs, or misses wall-clock ceiling repeatedly. This is a [[self-healing-loop]] trigger: first attempt recovery via POSTMORTEM + prompt fix; if a third run still underperforms, retire and encode hard-won lessons into the successor's PROMPT-BRAIN. ### (d) The surface it inte