← Back to catalog

tool

ASM — Agent Skill Manager Cheatsheet

Checklist-first guide to install, audit, organize, evaluate, develop, and publish portable AI agent skills with agent-skill-manager, the universal TUI and CLI for skill workflows.

  • asm
  • agent-skill-manager
  • skills
  • cli
  • tui
  • security
  • eval
  • registry
  • bundles

ASM — Agent Skill Manager Cheatsheet

One-line: ASM (agent-skill-manager) is the universal TUI and CLI for finding, installing, auditing, deduplicating, evaluating, developing, and publishing AI agent skills across Claude Code, Codex, Pi, OpenCode, Cursor, Windsurf, and more.

Installation

Install from the official npm package when possible. ASM requires Node.js >=18 <23 and npm >=9.

npm install -g agent-skill-manager
asm --version

Official one-liner install:

curl -sSL https://raw.githubusercontent.com/luongnv89/agent-skill-manager/main/install.sh | bash
asm --version

Launch the interactive dashboard:

asm

If a fresh install appears stale, check for shadowed binaries:

asm doctor
which -a asm

Step-by-Step Setup & Optimization

Step 1 — Inventory your skill estate

Start by seeing what ASM can discover across global and project scopes.

asm list
asm stats
asm search "code review"
asm inspect my-skill

Use filters when the inventory is large:

asm list --scope global --sort location
asm list --summary
asm list --compact
asm list --group-by tool
asm list --limit 20
asm search "code review" --json

Use the TUI (asm) for exploration and the CLI for repeatable commands.

Step 2 — Configure providers and scopes

ASM ships with 18 built-in providers enabled by default, including Claude Code, Codex, OpenClaw, Agents, Cursor, Windsurf, Cline, Roo Code, Continue, GitHub Copilot, Aider, OpenCode, Zed, Augment, Amp, Gemini CLI, Google Antigravity, and Hermes.

Open the config when you want to disable providers or add custom paths:

asm config show
asm config path
asm config edit

Use provider and scope flags to target exactly where a command applies:

asm list --tool claude --scope global
asm search "frontend" --tool codex --scope both
asm inspect my-skill --json

Step 3 — Install skills safely

Install from the ASM Registry by bare or scoped name:

asm install code-review
asm install luongnv89/code-review
asm install code-review --no-cache

Install from GitHub when the skill is not in the registry:

asm install github:user/my-skill
asm install github:user/my-skill#v1.0.0 -p claude
asm install github:user/skills --path skills/code-review
asm install github:user/skills --all -p claude -y
asm install https://github.com/user/skills/tree/main/skills/agent-config
asm install github:user/skills#main:skills/agent-config

For private repositories, use SSH transport:

asm install github:user/private-skill --transport ssh
asm install github:user/private-skill --transport auto

Preview risk before install when you do not fully trust the source:

asm audit security github:user/repo
asm install github:user/repo -p claude --yes --json

Step 4 — Clean, disable, update, and export

Find duplicates and remove redundant installs:

asm audit
asm audit --yes

Temporarily remove skills from agent discovery without deleting them:

asm disable my-skill
asm disable 'workflow*' --tool claude
asm enable my-skill

Keep installed skills current after reviewing changes:

asm outdated
asm update
asm update my-skill

Remove skills intentionally:

asm uninstall old-skill
asm uninstall old-skill --yes

Back up and restore your skill inventory:

asm export > skills-manifest.json
asm import skills-manifest.json

Step 5 — Develop skills with a live local loop

Scaffold a new skill:

asm init my-skill
asm init my-skill -p claude
asm init my-skill --path ./skills

Link local skills into an agent for live development. Because asm link uses symlinks, edits to the source folder are visible immediately.

asm link ./my-skill -p claude
asm link ./my-skill -p codex
asm link ./skill-a ./skill-b ./skill-c -p claude
asm link ./my-skills-folder -p claude
asm link ./my-skill --name my-alias -p claude
asm link ./my-skill -p claude --force

Validate the install path as an end user would see it:

asm install github:you/awesome-skill
asm install github:you/awesome-skill -p claude
asm install github:you/skills --path skills/awesome-skill
asm install github:you/awesome-skill --force
asm install github:you/awesome-skill -p claude --yes --json

Step 6 — Audit and evaluate skill quality

Run a security audit before publishing or installing unknown skills:

asm audit security my-skill
asm audit security ./path/to/my-skill
asm audit security --all

Run static quality evaluation with concrete improvement suggestions:

asm eval ./my-skill
asm eval ./my-skill --machine
asm eval ./my-skill --fix
asm eval-providers list

Verification checks basic index eligibility: valid frontmatter, meaningful body, no malicious patterns, and a readable skill directory with SKILL.md.

Quality evaluation goes deeper: structure, frontmatter, clarity, prompt engineering, context efficiency, safety, testability, and naming.

Step 7 — Publish and distribute

Publish a GitHub-hosted skill to the ASM Registry so users can install it by name:

asm publish ./my-skill
asm publish --dry-run ./my-skill
asm publish --force ./my-skill
asm publish ./my-skill --yes --machine

The publish pipeline validates frontmatter, runs security audit, generates a manifest with commit SHA and skill path, then opens a PR against the ASM Registry through gh.

Use bundles when a workflow needs a curated set of skills:

asm bundle list --predefined
asm bundle install frontend-dev
asm bundle install ./my-bundle.json
asm bundle create my-workflow
asm bundle export my-workflow ./my-workflow.json
asm bundle show my-workflow
asm bundle remove my-workflow

Step 8 — Automate with JSON and machine output

Use JSON for ad-hoc scripts and --machine for stable CI envelopes.

asm list --json
asm search "security" --json
asm inspect my-skill --json
asm eval ./my-skill --machine
asm publish ./my-skill --dry-run --machine

Use command help as the source of truth for flags on your installed version:

asm --help
asm install --help
asm audit security --help
asm eval --help

Best Practices

  • Use asm TUI for discovery, then copy the exact CLI command into scripts once the workflow stabilizes.
  • Run asm audit security before installing from unfamiliar GitHub repos; skills can instruct agents to execute code.
  • Prefer registry names for normal installs; prefer pinned Git refs for reproducible private or team workflows.
  • Use asm link for local development and asm install to test the clean end-user path.
  • Run asm eval before publishing; use --fix only after reviewing the proposed deterministic changes.
  • Keep provider scope explicit in destructive commands: combine --tool, --scope, and --yes carefully.
  • Use asm disable for conflicting or noisy skills before deleting them permanently.
  • Commit and review generated skills like code: frontmatter, scripts, references, and security posture all matter.
  • Use --json or --machine for CI and dashboards instead of parsing colored terminal output.
  • Run asm doctor when installs, PATH resolution, or provider discovery looks wrong.

Quick Command Reference

GoalCommand
Installnpm install -g agent-skill-manager
One-liner installcurl -sSL https://raw.githubusercontent.com/luongnv89/agent-skill-manager/main/install.sh | bash
Launch TUIasm
Versionasm --version
Health checkasm doctor
List skillsasm list
Search skillsasm search <query>
Inspect skillasm inspect <skill-name>
Install from registryasm install code-review
Install scoped registry skillasm install author/skill
Install from GitHubasm install github:user/repo
Install subfolder skillasm install github:user/repo --path skills/name
Install all repo skillsasm install github:user/repo --all -y
Security auditasm audit security <name-or-source>
Duplicate auditasm audit
Disable skillasm disable <target>
Enable skillasm enable <target>
Update skillsasm update [name...]
Show outdated skillsasm outdated
Remove skillasm uninstall <skill-name>
Export inventoryasm export > skills.json
Import inventoryasm import skills.json
Scaffold skillasm init my-skill
Link local skillasm link ./my-skill -p claude
Evaluate qualityasm eval ./my-skill
Auto-fix quality issuesasm eval ./my-skill --fix
List eval providersasm eval-providers list
Publish skillasm publish ./my-skill
Dry-run publishasm publish --dry-run ./my-skill
List bundlesasm bundle list --predefined
Install bundleasm bundle install frontend-dev
Show configasm config show
Edit configasm config edit
Stable CI outputasm eval ./my-skill --machine

Expected Outcomes

  • A single inventory of installed and available skills across supported AI agents.
  • Fewer duplicate or conflicting skills after asm audit, asm disable, and targeted uninstall/update workflows.
  • Safer installs because GitHub sources and local skills can be security-audited before use.
  • Faster skill authoring through asm init, asm link, asm inspect, and asm eval.
  • Registry-ready skills with verified frontmatter, meaningful instructions, audit results, and publish manifests.
  • Scriptable automation through --json and --machine output.

Reference

Sources & deeper reading

Official sources

Broader skill-quality context