Skip to content
Back to skills

Reflect

ASecurity

Self-improvement pass — read recent episodes, lessons and proposals, find repeated friction, repeated manual steps and checks done by eye, and propose concrete improvements (a skill, rule, gate or ADR) in memory/PROPOSALS.md. Creates nothing without the owner's yes. Use every few sessions, when the owner asks what to improve, or right after a repeated mistake.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 22, 2026
ai-agents

Security analysis

A100/100

Scanned October 22, 2026

npx -y skills add tornikebolokadze1-cyber/agent-foundation-kit --skill reflect --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Reflect?

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

Security grade badge for Reflect
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/tornikebolokadze1-cyber-reflect/badge)](https://www.skillsdirectory.com/skills/tornikebolokadze1-cyber-reflect)

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: reflect
description: Self-improvement pass — read recent episodes, lessons and proposals, find repeated friction, repeated manual steps and checks done by eye, and propose concrete improvements (a skill, rule, gate or ADR) in memory/PROPOSALS.md. Creates nothing without the owner's yes. Use every few sessions, when the owner asks what to improve, or right after a repeated mistake.
argument-hint: "[optional focus]"
---

# /reflect — the project improves itself, with the owner's consent

1. **Read** `memory/episodes/INDEX.md` and the last ~5 episodes (their "გაკვეთილები" and
   "წინადადებები" sections), `memory/LESSONS.md`, `memory/PROPOSALS.md`, and the latest gate/drill output
   if relevant.
2. **Find patterns** — the same mistake twice (→ lesson + a gate that prevents it); the same multi-step
   instruction given by the owner twice (→ skill candidate); a check done by eye that a script could do
   (→ gate candidate); a decision made implicitly (→ ADR candidate); a stale fact in CLAUDE.md, AGENTS.md
   or docs (→ correction).
3. **Write each as a proposal** in `memory/PROPOSALS.md`: `- [ ] **P-NNN** <change> — why (episode
   links as evidence) — cost/risk — how we will know it worked (a machine check if possible)`.
4. **Present the top 3** to the owner in Georgian, one sentence each, with your recommendation.
5. **Only after an explicit yes:** implement it; for a new skill use `.claude/skills/<name>/SKILL.md`
   with a name that does not collide with `~/.claude/skills` (global skills win on a clash) and a
   description that says when to use it; add a drill for any script. Then mark the proposal `[x]` with
   the date and move it under "გადაწყვეტილი". A "no" is recorded there too, with the reason.

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.