agentleFS
freskhufreskhu/genesis/.claude/skills/twin-update/SKILL.md

twin-update

Skill5 starsChanged 6 months agoApache-2.0View on GitHub

Detect new info about the owner from the session and update Owner Digital Twin files

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 11 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Check MemPalace for recent user-relevant signal:

    line 18
  2. Distinctive:

    If nothing new was detected, skip to Step 4 (log and exit).

    line 39
  3. Distinctive:

    Only update when genuinely new info surfaces

    line 63
  4. Distinctive:

    Keep updates concise — 1-2 lines per new fact

    line 64
  5. Distinctive:

    Use the same language and tone as existing content

    line 65
  6. Distinctive:

    If uncertain whether something is worth adding, err on the side of NOT adding

    line 66
  7. Distinctive:

    Always log, even if no updates were made:

    line 70
  8. Distinctive:

    Don't add trivial info ("the owner asked about ski boots" is NOT a twin update)

    line 93
  9. Distinctive:

    Don't duplicate what's already in MemPalace — the twin is a curated summary, not a log

    line 94
  10. Distinctive:

    Don't add speculative info — only confirmed facts or explicit preferences

    line 95
  11. Distinctive:

    Don't rewrite entire files — surgical edits only

    line 96

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Owner Digital Twin Update

When to Run

  • End of every session — The orchestrator checks if new info about the owner surfaced
  • After adding user/feedback knowledge entries — may contain preference changes
  • Manually — when the owner shares personal/professional updates

Step 1 — Gather Signal

Check MemPalace for recent user-relevant signal:

python3 scripts/palace.py search "<owner> preferences feedback" --wing owner --limit 10
python3 scripts/palace.py search "<owner>" --hall hall_preferences --limit 10

Cross-check against the last run — only signal newer than this date matters:

PRAGMA foreign_keys = ON; PRAGMA journal_mode = WAL; PRAGMA busy_timeout = 5000;
SELECT MAX(occurred_at) FROM activity_history WHERE action = 'twin_update_completed';

Also consider: what did the owner reveal during this session?

  • New interests or projects mentioned?
  • Preference corrections ("don't do X", "I prefer Y")?
  • Career or business updates?
  • New tools, technologies, or areas of focus?
  • Changes in communication style preference?

If nothing new was detected, skip to Step 4 (log and exit).

Step 2 — Classify and Route

Map each piece of new info to the correct file:

Info TypeTarget File
Career, role, company, educationOwner/identidade.md
Work style, communication, what to avoid/repeatOwner/preferencias.md
New interests, areas of focus, hobbies, techOwner/interesses.md
Brand positioning, narrative, public personaOwner/marca-pessoal.md

Step 3 — Update Files

For each file that needs updating:

  1. Read the current file — always read before editing
  2. Check for duplicates — don't add info that's already there
  3. Edit surgically — add or modify only the relevant section
  4. Keep the format — match existing style, concise, factual, PT-PT
  5. Never remove existing info unless it's explicitly contradicted

Rules:

  • Only update when genuinely new info surfaces
  • Keep updates concise — 1-2 lines per new fact
  • Use the same language and tone as existing content
  • If uncertain whether something is worth adding, err on the side of NOT adding

Step 4 — Log

Always log, even if no updates were made:

INSERT INTO activity_history (actor_id, action, entity_type, summary, metadata, occurred_at)
VALUES (
  1,
  'twin_update_completed',
  'owner_twin',
  CASE WHEN {updates_made} > 0
    THEN 'Digital Twin updated: ' || {files_changed}
    ELSE 'Digital Twin checked — no updates needed'
  END,
  json_object(
    'files_updated', json_array({list_of_files}),
    'entries_checked', {count},
    'updates_made', {updates_made}
  ),
  strftime('%Y-%m-%dT%H:%M:%SZ', 'now')
);

Anti-Patterns

  • Don't add trivial info ("the owner asked about ski boots" is NOT a twin update)
  • Don't duplicate what's already in MemPalace — the twin is a curated summary, not a log
  • Don't add speculative info — only confirmed facts or explicit preferences
  • Don't rewrite entire files — surgical edits only

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.