Skip to content
Back to skills

Hoo Reflector

ASecurity

Use to distill lessons from recent signals and outcomes - "run the reflector", weekly routine step 4, or "what did we learn". Proposes skillbook deltas; the curator merge is gated.

  • 5 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added September 6, 2026
ai-agentspythongit

Security analysis

A100/100

Scanned September 6, 2026

npx -y skills add shalintripathi/organic-os --skill hoo-reflector --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Hoo Reflector?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Hoo Reflector
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/shalintripathi-hoo-reflector/badge)](https://www.skillsdirectory.com/skills/shalintripathi-hoo-reflector)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: hoo-reflector
description: Use to distill lessons from recent signals and outcomes - "run the reflector", weekly routine step 4, or "what did we learn". Proposes skillbook deltas; the curator merge is gated.
---

# Reflector (weekly) - propose deltas, never write directly

1. Read: all signals/ since the last reflection, outcomes/ records, current
   skillbook.md, the last reflections/ file.
2. Propose deltas, each referencing an entry ID or NEW:
   - NEW lesson (with evidence tag + source: which signal/outcome proves it)
   - `S-nnn helpful+1` / `harmful+1` (an outcome confirmed/contradicted it)
   - `S-nnn deprecate` (contradicted repeatedly - cite the evidence)
   - `S-nnn edit: <new text>` (sharpen wording; meaning preserved)
   NEVER propose a full rewrite. NEVER propose deleting lines.
3. Write reflections/YYYY-Www.md listing every delta + evidence.
4. Stale review: list the entries whose last-confirmed date has aged past
   their evidence tier's threshold.
   `PYTHONPATH="$CLAUDE_PLUGIN_ROOT/lib" python3 -c "..."` snippet calling
   `core.contracts.skillbook_stale(<brain>)`, which returns
   `{id, evidence, last_confirmed, days_stale, threshold, text}` per entry,
   most stale first. Thresholds by tier: anecdotal 90 days, moderate 180,
   strong 365 (canonical in `core.contracts.STALE_DEFAULTS`; a site can
   override any tier with `skillbook: {stale_days: {...}}` in
   site-profile.yaml). Deprecated entries never appear here.
   Present each one for re-validation next to the evidence from this
   week's signals and outcomes that would confirm or contradict it:
   - Re-confirmed (the lesson still holds): propose it as a delta applied
     through `skillbook_update(root, "S-nnn")`, which stamps
     last-confirmed to today. That refresh is the point of the review.
   - Cannot be re-confirmed: PROPOSE deprecation as a delta, citing what
     is missing. The reflector NEVER auto-deprecates and never writes the
     book on its own - a stale entry is a question for a human, not a
     verdict, exactly like every other delta in step 2.
   An entry nobody can speak to this week is left as it is and listed
   again next week. Age is a prompt to re-check, never a reason to delete.
5. Curator merge: if the profile marks curator as gated (default), present the
   deltas for approval (in-session or channel); on approval apply each via
   `skillbook_append` / `skillbook_update` calls - one call per delta.
6. Commit "reflection: YYYY-Www" if git.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.