← ClaudeAtlas

post-itlisted

Use when the acting agent needs to store, retrieve, resolve, hand off, or promote a scoped working-memory item in support of current work. Not for primary task routing, broad artifact triage, or replacing the capability that owns the current work.
dmlguq456/hearting · ★ 7 · AI & Automation · score 78
Install: claude install-skill dmlguq456/hearting
# post-it ## Purpose Maintain a user-controlled working-memory surface distinct from automatic project memory under `<agent-home>/projects/*/memory/`. The primary mutation path is an explicit `post-it` action. The acting agent may also record a note when the proactive-nudge contract in [references/nudge-and-boundaries.md](references/nudge-and-boundaries.md) clearly applies; this is a judgment based on task relevance, not a fixed language signal. Treat a post-it as temporary. Durable truth belongs in source artifacts such as `plans/`, `documents/`, `spec/`, code, git history, and structured profile records. Use post-its only to bridge work that still needs to remain salient across sessions, then remove or promote them when their information graduates. ## Invariants - The user should not need to inspect post-it records line by line. Keeping the store lean and pruning graduated records are agent responsibilities. - During automatic nudges, remove only clearly graduated or stale records and report one concise summary; keep ambiguous records. - Ask only whether an action should be stored when confirmation is required. Show line-level sweep previews only for an explicit `post-it sweep` request. - Every entry eventually graduates or expires; do not build a permanent unstructured archive. ## Unified Store Post-its use the project-scoped working tier in the unified [memory store](../../tools/memory/README.md), backed by `memory.db` in SQLite WAL mode. - write working records w