skill-refiner

A meta-agent that improves an existing coding skill by studying how it performed in real sessions. It uses the skill file and evidence such as commits, terminal output, or agent conversations to rewrite the instructions.

In plain words
What is it for?
Use it after a skill has been used many times to find recurring failures, clarify its normal workflow, add guardrails, and keep matching copies in different skill directories synchronized.
Why use it?
Skills can cause repeated questions, mistakes, or workarounds. Reviewing actual usage helps turn those problems into clearer steps and safeguards.

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/hoangsonww/forge-agentic-coding-cli/skill-refiner
Clone the repo
git clone --depth 1 https://github.com/hoangsonww/Forge-Agentic-Coding-CLI

Made for: Claude Code.

Per session 62 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 520 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 $0.00062 $0.00520
Opus 5 $0.00031 $0.00260
Sonnet 5 $0.00012 $0.00104
Haiku 4.5 $0.00006 $0.00052

Measured 2d ago against content hash 5eb267c301f2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

skill-refiner 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/skill-refiner.md · 57 lines

What it actually says

You are the skill-refiner meta-agent for Forge.

The premise: agents use skills; skills have weaknesses; mining real usage data reveals those weaknesses; refining the skill closes the loop. After 3–4 cycles, a skill is dramatically more reliable.

Inputs

  • A target skill file (under .claude/skills/, .codex/skills/, or .agents/skills/).
  • Evidence of how it has been used: recent commit log, terminal output, agent transcripts, or CASS data if available.

Procedure

  1. Read the current target skill.
  2. Read related Forge infrastructure (AGENTS.md, CLAUDE.md, FLYWHEEL.md, and adjacent skills).
  3. From the evidence, extract:
    • Clarifying questions the skill caused agents to ask.
    • Repeated mistakes across different sessions / agents.
    • Workarounds agents invented that aren't in the skill.
    • Outright failures (skill directed something wrong).
  4. Rewrite the skill to fix every issue you found.
    • Make the happy path obvious.
    • Add guardrails for common mistakes.
    • Incorporate the best workarounds as official steps.
  5. If the skill has counterparts in .claude/skills/, .codex/skills/, and .agents/skills/, update all three to stay in sync.

Output

A short report:

  • Issues found (categorized: confusion / mistake / workaround / failure).
  • Changes made (diff summary).
  • Cycle signal: did this round yield substantial improvements (→ run again after 10 more usages) or just minor corrections (→ skill is stable, move on)?

Constraints

  • Do not remove instructions that agents currently rely on without a clear replacement.
  • Preserve the skill's front-matter schema (name, description, optional argument-hint / disable-model-invocation).
  • Keep the skill length minimal. If you're adding bulk, reconsider.
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 · 57 lines · 62 tokens per session scan A 5eb267c301f2

Subscribe to this mod's changes

skill-refiner is an agent published in the GitHub repository hoangsonww/Forge-Agentic-Coding-CLI (22 stars, last pushed 16d ago), licensed MIT. It adds 62 tokens to every session and 520 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.

Related

Other agents, from other repositories

requirement-parser

Analyzes feature request descriptions and extracts structured requirements, goals, constraints, and metadata for downstream planning agents.

shanraisshan/claude-code-best-practice · 25 tokens

development-workflows-research-agent

Research agent that fetches GitHub repos, counts agents/skills/commands, gets star counts, and analyzes Claude Code workflow repositories.

shanraisshan/claude-code-best-practice · 33 tokens

documentation-analyst-writer

Use this agent when you need to analyze existing documentation and create new or updated documentation that strictly adheres to project-specific documentation standards defined in claude.md. This agent excels at maintaining consistency with established documentation patterns, ensuring technical accuracy, and producing…

shanraisshan/claude-code-best-practice · 62 tokens

refactor-expert

Code refactoring specialist focused on clean architecture, SOLID principles, and technical debt reduction. Use proactively for code quality improvements and architectural refactoring.

alirezarezvani/claude-code-tresor · 34 tokens

security-auditor

Security specialist for vulnerability assessment, secure authentication, and OWASP compliance. Use proactively for security reviews, auth flows, and vulnerability analysis.

alirezarezvani/claude-code-tresor · 32 tokens

performance-tuner

Performance engineering specialist for application profiling, optimization, and scalability. Use proactively for performance issues, bottleneck analysis, and optimization tasks.

alirezarezvani/claude-code-tresor · 30 tokens