meta-improve

meta-improve is a skill for Claude Code from wrg32786/aigent-os. It costs 27 tokens per session (483 once invoked), scanned A, original, MIT.

A controlled self-improvement workflow that applies an already approved change proposal to an AI project after checking its evidence and tests.

In plain words
What is it for?
Use it to implement one operator-approved improvement with traceable evidence and verification.
Why use it?
It prevents an AI system from changing itself based only on an unapproved suggestion and requires proof before and after the change.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter. Also seen: agent in frontmatter.

Needs its repository: it runs a file that does not travel with it, so clone the repository first. The line is node daemons/nightly-contracts.mjs meta --root <aigent-root> \.

Good fit Use it to implement one operator-approved improvement with traceable evidence and verification.

Compare 6 skills from other repositories ↓
Install

Getting it into your agent

It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.

Clone the repo
git clone --depth 1 https://github.com/wrg32786/aigent-os
agentmods
npx agentmods add skills/wrg32786/aigent-os/meta-improve

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for meta-improve

README.md
[![agentmods](https://agentmods.dev/badge/skills/wrg32786/aigent-os/meta-improve.svg)](https://agentmods.dev/skills/wrg32786/aigent-os/meta-improve)
Your own site
<a href="https://agentmods.dev/skills/wrg32786/aigent-os/meta-improve"><img src="https://agentmods.dev/badge/skills/wrg32786/aigent-os/meta-improve.svg" alt="Measured on agentmods" height="20"></a>
Per session 27 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 483 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
SkillSpector: 1 finding, up to low

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

  • low Excessive Agency · line 49
    Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.
    Fix: Limit the skill's scope to its documented purpose. Remove instructions that enable the agent to perform actions outside its stated functionality.
How audits are shown
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 5.1 $0.00027 $0.00483
Opus 5 $0.00014 $0.00242
Sonnet 5 $0.00005 $0.00097
Haiku 4.5 $0.00003 $0.00048

Measured 7d ago against content hash c124ef535f74, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

meta-improve 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 7d 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.

skills/meta-improve/SKILL.md · 65 lines

What it actually says

Constrained self-improvement

The public verb is /meta-improve-vault.

Input and gate

Select only from canonical vault/memory/DREAM_LOG.md: an exact candidate whose status is approved and whose approval was written by /dream --set-status from a direct operator decision. proposed is not approval.

Read only the exact evidence references named by that approved candidate from vault/memory/runtime/LESSONS.jsonl and vault/memory/runtime/NIGHTLY_CAPTURE_CANDIDATES.jsonl. Those products provide traceability; they never provide approval by themselves.

Before Build:

node daemons/nightly-contracts.mjs meta --root <aigent-root> \
  --candidate <candidate-id>

Require META_GATE PASS. It checks the canonical block for exact approved status and non-placeholder operator evidence. This is the executable nightly-contracts.mjs meta gate.

Build

For one approved candidate:

  1. Re-read its evidence, proposed change, mutation proof, and scope.
  2. Use repository isolation where normal policy requires it.
  3. Apply the smallest change satisfying the proposal.
  4. Run the real red-before and green-after proof plus relevant system checks.
  5. Produce a review artifact with exact diff, commands, exits, residual risks, and rollback.
  6. Surface accept / reject / revise to the operator.

Never merge, push, deploy, broaden scope, modify an operator gate, approve another candidate, or infer approval. A failed proof asks the Dream writer to set the candidate blocked.

Status ownership

This skill never edits Dream directly. The Dream writer owns:

  • direct operator approval: proposed -> approved;
  • verified build failure: approved -> blocked;
  • direct rejection: approved -> rejected;
  • independently verified merge: approved -> merged.

The complete flow is:

propose -> operator gate -> build -> independent review -> operator outcome

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. 7d ago First seen · 65 lines · 27 tokens per session scan A c124ef535f74

Subscribe to this mod's changes

meta-improve is a skill published in the GitHub repository wrg32786/aigent-os (18 stars, last pushed today), licensed MIT. It adds 27 tokens to every session and 483 once invoked, about $0.0001 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 2026-09-01.