managing-memorylisted

# Memory A thin layer over Claude Code's native CLAUDE.md. We don't reinvent persistence — Claude Code already loads `CLAUDE.md` (project / user / local / managed) automatically at session start. What this skill adds: 1. **Auto-classification on `/llm-orchestrator:remember`** — facts route into `## Conventions`, `## Decisions`, `## People`, or `## Notes` sections of the right CLAUDE.md. 2. **Soft-delete via `/llm-orchestrator:forget`** — removed lines move to `~/.llm-orchestrator/memory/.trash/` so accidents are recoverable. 3. **Plugin-internal memory** at `~/.llm-orchestrator/memory/<project-hash>.md` for research-gate state that doesn't belong in user-facing CLAUDE.md — `## Research config` (aggressiveness knob) and `declined_mcp:` entries. Two native surfaces, two owners: **CLAUDE.md** is the *user's* curated memory — that's what this skill writes to. The harness may also keep an *assistant-owned* auto-memory directory (Claude Code persists the model's own notes across sessions); this skill never writes there, and facts the user asks to remember belong in CLAUDE.md, not in assistant notes. ## Where things live | Target file | What goes there | Loaded by | |-------------------------------------------------|-------------------------------------------------------------|----------------------------------------| | `./CLAUDE.md` (project) | P