Back to directory

agent-skill-manager

A scriptable CLI that installs, searches, audits and organizes AI agent skills across Claude Code, Codex, Cursor and 18 other tools, tracking token cost and letting a skill be read without installing it.

agent-skill-manager

agent-skill-manager (asm) addresses skill sprawl across AI coding tools: the same skill copied with different versions into separate tool-specific directories, with no single inventory of what is installed, duplicated or outdated. Every command supports --json and --yes for non-interactive, agent-driven use, and an optional terminal UI supports local browsing.

asm list gives one view across all 21 supported providers and scopes; asm search, asm inspect, asm stats and asm audit cover discovery and inventory; asm stats --tokens and asm audit residency measure the context-token cost of an installed skill's description being resident in every prompt whether or not it fires; and asm get prints a skill's body without installing it, for reading a skill once rather than keeping it resident forever. Installing with asm install validates a skill's frontmatter, scans for shell execution, network access and credential exposure, and pins the registry commit installed.

Its catalog covers 6,082 skills from 77 repositories, browsable online, and it is MIT licensed with no accounts or telemetry.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.