Core artifact map and lifecycle rules for `signal` (deferred/unverified attention), `task` (committed work with an outcome), and `learning` (scoped belief from evidence). Briefly covers note, mission, brief, person, and skill-proposal routing. Use for choosing a core kind, writing signal/task bodies, promoting signal->task, and closing tasks with `## Outcome`. Usually read after [[proactive-capture/SKILL.md]] routes you here.
Use on daily cadence, at session start, or when the user says "what's due today" / "what am I forgetting" / "daily check-in". Reads today's due `check_at` items, classifies each (silent resolve / surface / defer), and brings the user only what needs attention. Every surfaced item must close before session ends.
Use on monthly cadence or when the user asks "is this still the right mission?" / "should we reset?" / "anything structural worth locking in?". Detects two drifts - (1) mission drift: active `mission` no longer matches recent tasks; (2) structural drift: recurring usage that should crystallize into a new `domain` or `kind`. Surfaces evidence; user decides.
Use when picking between note / new kind / new domain — before `loopany kind propose`, before extracting a new domain, or any time the user says "I want to track X" / "should we add a kind for…" / "let's make a domain for…". Default: write a note. Triggers on any new-concept decision in a loopany workspace.
Use immediately after substantive work concludes in a loopany-enabled session — PR shipped, incident resolved, decision made, problem noticed but not yet acted on, outcome observed. Routes to the right kind, then hands off to [[core-artifacts/SKILL.md]] (signal/task body discipline) or [[reflect/SKILL.md]] (learnings). Specifies the subagent dispatch pattern. Main session reads to filter; subagent reads to write.
Use when the user says "reflect" / "what have we learned" / "improve yourself", on weekly cadence, or after ≥3 `task`s flip to done in a short window. Reads recent outcomes + dismissed signals + rejected proposals, writes `learning` and (optionally) `skill-proposal` artifacts. Also use when writing a `learning` or `skill-proposal` from any source. The self-iteration loop.
Pick the right `relation` verb (led-to / addresses / mentions / supersedes / follows-up / cites) when running `loopany refs add`, writing `mentions[]` in frontmatter, or any `[[id]]` wiki link. Read before creating any reference edge — synonyms fragment the graph.
Use on weekly cadence or when the user says "what's slipping" / "weekly check" / "is the workspace healthy". Three problems it sweeps: drift (`loopany doctor`), slippage (overdue `check_at`), parking lots (running tasks ≥14d, active signals ≥7d, pending proposals). Forces closure on stuck artifacts. Distinct from the daily [[daily-followups/SKILL.md]].