| name | whoami |
| description | Show the auto-curated user profile (memory/USER.md) — who the user is, preferences, working patterns, recent corrections, and open threads — with last-updated, section counts, and a token-cap/prune check. Use when asked "what do you know about me", "show my profile", "whoami", or to review/prune the profile. |
| category | meta |
| tags | ["user-profile","memory","self-knowledge","review","curation"] |
| maturity | beta |
| external_deps | [] |
whoami
Renders the auto-curated user profile — the structured memory/USER.md that the
memory-curate cron job (scripts/memory-curate-cron.sh, every 2 h) maintains and that
session-init.sh loads into every session.
This is the read-only review surface for the profile (Tier 2 item 4C). It does not edit
anything — it shows the profile, how big it is against the ~500-token cap, and where it
needs pruning.
Run it
bash "$CLAUDE_SKILLS_DIR/whoami/scripts/whoami.sh"
If $CLAUDE_SKILLS_DIR is unset, the script lives at ~/.claude-agent/.claude/skills/whoami/.
What it shows
- The full profile, rendered.
- Last updated timestamp and size / token estimate vs the 500-token cap.
- Per-section counts across the five schema sections:
stable_facts · preferences · working_patterns · recent_corrections · open_threads.
- A prune nudge for the rolling sections (recent_corrections keeps newest 8,
open_threads keeps newest 6).
The schema (maintained by memory-curate.cjs)
- stable_facts — identity, timezone, environment, tooling (rarely change)
- preferences — how the user likes work done
- working_patterns — recurring workflows, cadences
- recent_corrections — explicit corrections / things that superseded an old belief (newest first)
- open_threads — in-flight work to resume
How to prune
This skill is read-only. To prune, edit memory/USER.md by hand:
- fold a settled
recent_corrections entry into preferences/stable_facts and delete the correction,
- remove
open_threads that are finished,
- tighten any
stable_facts/preferences that have drifted.
The distiller never deletes core facts — only you do, here. It only trims the rolling
sections to newest-N and enforces the token cap.
Rules
- Never edit the profile from this skill — it only reports.
- If the profile is over the token cap, surface that first and recommend a prune.
- CLAUDE.md is separate and hand-written —
whoami shows the learned profile, not the rules.
