task-observer: learn skill updates from 2026-06-29

.agents/skills/learn/SKILL.md

@@ -104,6 +104,8 @@ Real entries already in the canonical voice. Match this shape and rhythm — the
 
 Use the abstraction ladder from `docs/ai/learning-system.md`, but do not force a polished rule out of every single bug. Raw bugs are useful samples. When several samples point to the same class, cluster them into the highest still-actionable principle; when there is only one sample, capture it directly only if the prevention surface is already clear or the risk is high. If a claim says a workflow "should" do something, verify the actual trigger, prompt, hook, test, structural check, or code path that makes it happen.
 
+When a bug finding involves a shared contract — a model field, API shape, hook return type, cache key, fixture format, or type definition — explicitly ask: which contract changed, and which consumer classes depend on it (UI, API routes, tests, fixtures, docs)? Climb to the highest actionable principle, e.g. "when a shared contract changes, every consumer must be updated," and pair it with a candidate enforcement hook: schema scan, propagation test, shared builder, or checklist.
+
 After identifying the principle, note what would enforce it next time: a regression test, lint rule, schema/contract scan, shared helper, checklist, docs update, skill tweak, or automation.
 
 Store prevention work as one readable list, e.g. `Prevention artifacts: docs (required), test (required), skill (proposed)`. Required artifacts describe work needed to prevent the issue; proposed artifacts are worthwhile ideas to consider. The executor decides what is executable now and records completed, blocked, deferred, or follow-up status. Any required code, test, helper, skill, automation, architecture, or enforcement artifact requires TDD/review discipline. If the code surface does not exist yet, docs, skill, or nested-AGENTS updates may be the only executable prevention artifact for now.