Self Improve

What it does

Example prompts

• “/self-improve”

Workflow steps

6 steps, taken from the first numbered list in SKILL.md.

1. 1User Revision (highest signal) — User edited or rewrote AI output
2. 2User Feedback — User explicitly says "I liked X" or "Don't do Y"
3. 3Acceptance Pattern — User accepted output without changes
4. 4Error Recovery — Something failed and was fixed
5. 5Self-Critique — Agent reviews its own output and spots issues
6. 6After-Action Review — Post-goal structured reflection

What it can do on your machine

Read from SKILL.md and the folder at commit 47e9570. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  No scripts in the folder and no shell commands in SKILL.md (its code samples are json).

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  No URLs in SKILL.md.

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Self Improve loads about 1.7k tokens when it runs. Until then it costs about 34 tokens; SKILL.md has 510 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~34

When it runs · the whole SKILL.md, loaded when a task matches

~1.7k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from Ckokoski/AuthorAgent at commit 47e9570, republished under its MIT licence (© Ckokoski). 510 words, ~1,704 tokens.

name

self-improve

description

Continuous self-improvement loop — AuthorAgent learns from mistakes, successes, and user feedback to get better over time

author

AuthorAgent

version

1.0.0

triggers

self improve, improve yourself, learn from, what did you learn, improvement log, get better, lessons learned, self reflection, review performance

permissions

file:read, file:write

  ┌─────────────────────────────┐
  │                             │
  │    1. DO THE WORK           │
  │    (goal step, writing,     │
  │     research, etc.)         │
  │                             │
  └──────────┬──────────────────┘
             │
             ▼
  ┌─────────────────────────────┐
  │                             │
  │    2. OBSERVE RESULT        │
  │    Did the user accept it?  │
  │    Did they revise it?      │
  │    Did it trigger an error? │
  │    How long did it take?    │
  │                             │
  └──────────┬──────────────────┘
             │
             ▼
  ┌─────────────────────────────┐
  │                             │
  │    3. EXTRACT LESSON        │
  │    What specifically        │
  │    went right or wrong?     │
  │    What pattern emerges?    │
  │                             │
  └──────────┬──────────────────┘
             │
             ▼
  ┌─────────────────────────────┐
  │                             │
  │    4. STORE LESSON          │
  │    Write to learning log    │
  │    (workspace/memory/       │
  │     improvement-log.jsonl)  │
  │                             │
  └──────────┬──────────────────┘
             │
             ▼
  ┌─────────────────────────────┐
  │                             │
  │    5. APPLY LESSONS         │
  │    Before each new task,    │
  │    check the log for        │
  │    relevant lessons and     │
  │    adjust behavior          │
  │                             │
  └──────────┬──────────────────┘
             │
             └──────── back to step 1

{
  "id": "lesson-042",
  "timestamp": "2026-02-24T14:30:00Z",
  "category": "writing",
  "trigger": "user_revision",
  "context": "Chapter 3 of thriller project",
  "observation": "User rewrote all dialogue tags from creative tags to simple said/asked",
  "lesson": "This user strongly prefers invisible dialogue tags (said/asked). Do not use creative tags like 'exclaimed', 'muttered', 'hissed' unless the user specifically asks.",
  "confidence": 0.9,
  "applied_count": 0,
  "source": "user_feedback"
}

## Lessons Learned (Apply These)
- This user prefers invisible dialogue tags (said/asked). Confidence: 0.9
- For thriller pacing, keep chapters under 3000 words. Confidence: 0.85
- When researching, include at least 3 specific sources. Confidence: 0.7
- Use Gemini for planning tasks (faster, good enough). Confidence: 0.8

show improvement log

what did you learn from [project/goal]

clear lesson [id]

improvement stats

© Ckokoski, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in skills/core/self-improve of Ckokoski/AuthorAgent.

Open the folder on GitHubat commit 47e9570