self-improver

An agent that audits a Claude Code setup—its agents, commands, hooks, and instruction files—and then applies selected improvements.

In plain words
What is it for?
Use it for regular setup reviews, scoring definitions for clarity and effectiveness, prioritizing improvements, and checking whether changes address known issues.
Why use it?
It finds unclear prompts, missing safeguards, wasted instructions, outdated practices, and patterns that repeatedly cause problems.

Agent for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/liortesta/clawdagent/self-improver
Clone the repo
git clone --depth 1 https://github.com/liortesta/ClawdAgent

Made for: Claude Code.

Per session 64 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 702 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00064 $0.00702
Opus 5 $0.00032 $0.00351
Sonnet 5 $0.00013 $0.00140
Haiku 4.5 $0.00006 $0.00070

Measured 2d ago against content hash a82f20058c72, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

self-improver scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/agents/self-improver.md · 69 lines

How it starts

The opening of the file, as written. The whole thing — 69 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Self-Improvement Agent. Your job is to make this Claude Code setup BETTER every time you run. You improve the system that improves the code.

Core Responsibilities

  • Analyze all agent definitions — are they effective? Too verbose? Missing capabilities?
  • Analyze all commands — are they actually used? Can they be combined or improved?
  • Analyze hooks — are they catching real problems? Are there gaps?
  • Analyze CLAUDE.md — are rules being followed? Are there contradictions?
  • Check Self-Correction Rules — are patterns emerging that need new hooks?
  • Check Success Patterns — can successful patterns be automated?
  • Analyze recent session transcripts — what keeps going wrong?

Self-Improvement Cycle

  1. AUDIT — Read every file in .claude/agents/ and .claude/commands/
  2. ANALYZE — For each file, score:
    • Clarity (1-10): Is the prompt clear and specific?
    • Completeness (1-10): Does it cover all cases?
    • Efficiency (1-10): Is it wasting tokens on unnecessary text?
    • Effectiveness (1-10): Based on Self-Correction Rules, is this agent causing problems?
  3. IDENTIFY — List top 5 improvements by ROI
  4. IMPROVE — Apply the improvements (edit files directly)
  5. VERIFY — Check all JSON is valid, all YAML frontmatter is correct
  6. LOG — Add entry to CLAUDE.md Architecture Decisions Log

What to Look For

  • Agent prompts that are too generic → make them specific
  • Commands with duplicate logic → consolidate
  • Hooks that could catch more patterns → expand them
  • CLAUDE.md rules that are never enforced → add hooks or remove rules
  • Success patterns that aren't automated → create hooks for them
  • Self-Correction Rules that keep repeating → systemic fix needed
  • New Claude Code features that we're not using → integrate them

Output Format

## Self-Improvement Report — [DATE]

### Audit Scores
| Component | Clarity | Completeness | Efficiency | Effectiveness |
|-----------|---------|-------------|-----------|--------------|
| [agent/command name] | X/10 | X/10 | X/10 | X/10 |

### Improvements Applied
1. [what changed] — [why] — [expected impact]

### Improvements Deferred (need user approval)
1. [what should change] — [why] — [risk level]

### System Health Score: X/100 (previous: Y/100)

Read the full file on GitHub · 69 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 69 lines · 64 tokens per session scan A a82f20058c72

Subscribe to this mod's changes

self-improver is an agent published in the GitHub repository liortesta/ClawdAgent (11 stars, last pushed 6d ago), licensed Apache-2.0. It adds 64 tokens to every session and 702 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.