skillkit CLAUDE.md

A Python library for giving AI agents packaged areas of expertise they can discover and use when needed. It supports skill files, multiple sources, several agent frameworks, and different language models.

In plain words
What is it for?
Building agents that find skills from projects, personal folders, or plugins; load them progressively; run their scripts with controls; and connect them to frameworks such as LangChain or CrewAI.
Why use it?
It avoids putting every instruction into an agent’s context at the start, while adding checks that help prevent unsafe file access.

Instructions file

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 instructions/maxvaega/skillkit/claude-md
Clone the repo
git clone --depth 1 https://github.com/maxvaega/skillkit
Per session 4,749 This file is loaded in full into every session.
When invoked 4,749 The same file — it is already loaded in full.
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 $0.04749 $0.04749
Opus 5 $0.02374 $0.02374
Sonnet 5 $0.00950 $0.00950
Haiku 4.5 $0.00475 $0.00475

Measured 2d ago against content hash 3e68a30bb715, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

skillkit CLAUDE.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 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.md · 329 lines

How it starts

The opening of the file, as written. The whole thing — 329 lines — stays where its author put it; the contents beside it link to each section on GitHub.

skillkit

skillkit is a Python library that implements Anthropic's Agent Skills functionality, enabling LLM-powered agents to autonomously discover and utilize packaged expertise. The library provides:

  • Multi-source skill discovery from personal directories, project directories, and plugins
  • SKILL.md parsing with YAML frontmatter validation
  • Progressive disclosure pattern (metadata loading → on-demand content)
  • Framework integrations (LangChain, LlamaIndex, CrewAI, Haystack, Google ADK)
  • Security features (path traversal prevention, permission checks)
  • Model-agnostic design supporting Claude, GPT, Gemini, and open-source LLMs

Development Approach

This project follows a Vertical Slice MVP strategy to deliver working functionality quickly:

  • v0.1 (Released): Core functionality + LangChain integration (sync only)
  • v0.2 (Released): Async support + multi-source discovery + plugin integration
  • v0.3 (Released): Script execution with security controls
  • v1.0 (Planned): Additional framework integrations + production polish + comprehensive documentation + 90% test coverage

Current Focus (v0.4)

The v0.4 release implements advanced progressive disclosure with intelligent caching:

  1. LRU Content Cache: In-memory cache with configurable size (default: 100 entries)
  2. Mtime-based Invalidation: Automatic cache invalidation when SKILL.md files are modified
  3. Argument Normalization: Whitespace variations map to same cache entry for maximum efficiency
  4. Thread-Safe Concurrency: Per-skill asyncio locks enable safe parallel invocations
  5. Cache Management API: get_cache_stats(), clear_cache() for monitoring and control
  6. Performance: <1ms cache hits vs 10-25ms first invocation (up to 25x faster)
  7. Memory Efficient: ~2.1KB per cached entry, ~5MB cache overhead
  8. High Hit Rate: 80%+ cache hit rate achievable with typical usage patterns
  9. Script Integration: Script detection integrates with Level 2 caching lifecycle
  10. Backward Compatible: 100% compatible with v0.1/v0.2/v0.3 APIs

Read the full file on GitHub · 329 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 · 329 lines · 4,749 tokens per session scan A 3e68a30bb715

Subscribe to this mod's changes

skillkit CLAUDE.md is an instructions file published in the GitHub repository maxvaega/skillkit (147 stars, last pushed 7mo ago), licensed MIT. It adds 4,749 tokens to every session, about $0.0237 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 instructions, from other repositories