Stats
Actions
Tags
From claude-automemory
Use when the user asks about prior sessions, recently touched files, past work, or refers to "what we did" / "yesterday" / "earlier" — invokes the bundled recall CLI to answer from local transcripts before falling back to grep.
How this skill is triggered — by the user, by Claude, or both
Slash command
/claude-automemory:claude-automemoryThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
A SessionStart hook has already injected an "orient" snapshot at the start of this session: the 10 most recent files touched in this cwd and the 3 most recent sessions, with a "RUN FIRST" instruction block.
A SessionStart hook has already injected an "orient" snapshot at the start of this session: the 10 most recent files touched in this cwd and the 3 most recent sessions, with a "RUN FIRST" instruction block.
When the user asks anything that hints at prior work, prefer the bundled CLI before falling back to grep, find, or glob. The CLI is read-only over ~/.claude/projects/<cwd-encoded>/*.jsonl and costs ~50 tokens per call.
Run from Bash:
python3 "${CLAUDE_PLUGIN_ROOT}/bin/recall.py" <command> [args]
| Trigger phrase from user | Command |
|---|---|
| "what did I touch yesterday" | files --days 1 --json |
| "files modified last week" | files --days 7 --json |
| "recent sessions" | list --json --limit 5 |
| "search for X" / "find sessions about X" | search "X" --json |
| "show me session abc..." | show abc12345 --json |
| "what plans do I have" | plans --json |
| "is recall working" | health --json |
| "across all projects" | append --global |
--days N works on files, list, search, plans.--limit N works on files, list, plans.--global widens scope from current cwd to all projects.--json for machine-parseable output (default for the hook).If the CLI errors, continue silently — recall is a convenience, not a blocker. Fall back to grep/find only after recall has been tried and returned empty or errored.
~/.claude/.auto memory system at ~/.claude/projects/<cwd>/memory/ — that handles persistent user/feedback/project memories. This handles transcript history.Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub gabrielpc1190/claude-automemory --plugin claude-automemory