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.
npx agentmods add instructions/quantsquirrel/claude-forge-smith/agents-mdgit clone --depth 1 https://github.com/quantsquirrel/claude-forge-smithWrote 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.
[](https://agentmods.dev/instructions/quantsquirrel/claude-forge-smith/agents-md)<a href="https://agentmods.dev/instructions/quantsquirrel/claude-forge-smith/agents-md"><img src="https://agentmods.dev/badge/instructions/quantsquirrel/claude-forge-smith/agents-md.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00594 | $0.00594 |
| Opus 5 | $0.00297 | $0.00297 |
| Sonnet 5 | $0.00119 | $0.00119 |
| Haiku 4.5 | $0.00059 | $0.00059 |
Grade A, and why
claude-forge-smith AGENTS.md 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 5d 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.
How it starts
The opening of the file, as written. The whole thing — 61 lines — stays where its author put it; the contents beside it link to each section on GitHub.
claude-forge-smith
Purpose
TDD 기반 스킬 자동 업그레이드 플러그인. AI 에이전트 스킬을 자동으로 분석, 평가, 개선하며 테스트 주도 개발(TDD) 원칙을 활용한 안전한 자기 개선 루프를 제공합니다. Gödel Machines 이론에 영감을 받아, 통계적 신뢰구간 검증으로 유의미한 향상만 적용합니다.
Key Files
| File | Description |
|---|---|
README.md |
Project documentation with research background and usage |
.claude-plugin/plugin.json |
Plugin manifest for Claude Code integration |
Subdirectories
| Directory | Purpose |
|---|---|
.claude-plugin/ |
Plugin metadata and configuration (see .claude-plugin/AGENTS.md) |
.github/ |
GitHub workflows, issue templates, PR templates (see .github/AGENTS.md) |
config/ |
Environment configuration settings (see config/AGENTS.md) |
data/ |
Data storage for baselines, detections, golden packets (see data/AGENTS.md) |
docs/ |
Project documentation and theoretical foundations (see docs/AGENTS.md) |
examples/ |
Usage examples and tutorials (see examples/AGENTS.md) |
hooks/ |
Claude Code hooks for tool/session lifecycle (see hooks/AGENTS.md) |
scripts/ |
Automation and utility scripts (see scripts/AGENTS.md) |
skills/ |
Skill definitions with SKILL.md files (see skills/AGENTS.md) |
tests/ |
Test suite with unit, integration, and e2e tests (see tests/AGENTS.md) |
For AI Agents
Working In This Directory
- This is a Claude Code plugin for skill self-improvement
- Core workflow: Trial Branch → Evaluate (3x) → Statistical validation → Merge/Discard
- Always use
trial-branch.shfunctions for isolated experiments - Never modify
evaluator.mdorrubric.mdduring bootstrapping
Testing Requirements
- Each skill has pressure scenarios in its directory
- Use 3 independent evaluations for statistical confidence
- Verify confidence interval separation before merging improvements
Common Patterns
- Lazy Detection: Only scan skills on Write/Edit operations
- Evaluator Separation: Different agent evaluates than improves
- 95% CI: Use t-distribution for small samples (n=3)
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.
- 5d ago First seen · 61 lines · 594 tokens per session scan A 2177283098a2
claude-forge-smith AGENTS.md is an instructions file published in the GitHub repository quantsquirrel/claude-forge-smith (2 stars, last pushed 6mo ago), licensed MIT. It adds 594 tokens to every session, about $0.0030 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 2026-08-31.
Other instructions, from other repositories
godmode AGENTS.md
AGENTS.md instructions for arbazkhan971/godmode, covering agents.md — godmode for ai coding agents, core workflow: the godmode loop, key principles, authoring discipline and how to use skills.
cms-cultivator CLAUDE.md
Claude Code instructions for kanopi/cms-cultivator, covering claude context for cms cultivator development, project overview, architecture: "agents orchestrate, skills guide", key architectural decisions and architecture: agents + skills.
cms-cultivator AGENTS.md
AGENTS.md instructions for kanopi/cms-cultivator, covering agent instructions, landing the plane (session completion) and quality gates.
claude-sports-analytics CLAUDE.md
Instructions for PuckAPI/claude-sports-analytics, covering hockey skills -- claude code instructions, public repo rules, never include, commit messages and code and content.
Eureka GEMINI.md
Instructions for jeonnoin-alt/Eureka, a project described as: Scientific TDD for AI coding agents: pre-registration before data, claims audit before submission, external novelty gate, reproducibility. Claude Code + 4 platforms.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.