mcpField
M
managing-skills

managing-skills

github.com/mikekellyAgent workflows271 installsUpdated 40w ago
Repository

About managing-skills

managing-skills is an agent skill for operating the `npx skills` CLI to install, discover, check, update, and list skills. It selects the agent target and defaults to project-level installation unless the user specifies otherwise.

Key features

  1. 01Install skills from repositories or paths
  2. 02Search for available skills
  3. 03Check for skill updates
  4. 04Update installed skills
  5. 05List installed skills

Use cases

  1. 01Add a skill to the current project
  2. 02Find skills matching a keyword
  3. 03Check or update installed skills

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/mikekelly/managing-skills --skill managing-skills

After installing, type/managing-skillsin Claude Code to confirm it loaded.

Bundled filesREADME.md

Details

From the project README
Manage agent skills via the `npx skills` CLI. Handle installing skills from GitHub repos, searching for available skills, checking for updates, and updating installed skills.

<quick_start> Determine which operation the user wants and run the appropriate command. Always include --yes to skip confirmations. Default to project-level install and the current agent type unless told otherwise.

Primary commands:

npx skills add {source} --yes --agent {agent-type}
npx skills find {keyword} --yes
npx skills check --yes
npx skills update --yes
npx skills add --list --yes

</quick_start>

Detect the agent you are running as. Map to the correct `--agent` flag value:
AgentFlag value
Ampamp
Claude Codeclaude-code
Clinecline
Codexcodex
Continuecontinue
Cursorcursor
Gemini CLIgemini-cli
GitHub Copilotgithub-copilot
Goosegoose
Kilo Codekilo
Kiro CLIkiro-cli
OpenCodeopencode
Qwen Codeqwen-code
Roo Coderoo
Traetrae
Windsurfwindsurf

If unsure, check for config directories (e.g., .claude/, .codex/, .cursor/). Only include additional agent types if the user explicitly requests it (e.g., "install for all agents" or "also install for codex"). </agent_type>

<install_scope>

  • Project (default): Installs to ./<agent>/skills/ in the current project.
  • Global (-g): Installs to ~/<agent>/skills/. Only use when the user says "global", "globally", or "for all projects". </install_scope>
User says: install, add, get, set up a skill 1. Identify the skill source. Accepts: - `owner/repo` — installs all skills from the repo - Full GitHub/GitLab URL to a repo, directory, or SKILL.md file (e.g., `https://github.com/owner/repo/tree/main/skills/foo`) - Local filesystem path - Use `-s skill-name` to cherry-pick a specific skill by name from a multi-skill repo 2. Determine scope: project (default) or global (`-g`). 3. Determine agent type(s) to target. 4. Run: ```bash npx skills add {source} --yes --agent {agent-type} ``` Add `-g` if global. Add multiple `--agent` flags if targeting multiple agents. "Install the vercel-labs/skills skill" → `npx skills add vercel-labs/skills --yes --agent claude-code`

"Install just the managing-skills skill from that repo" → npx skills add vercel-labs/skills --yes -s managing-skills --agent claude-code

"Install this skill: github.com/owner/repo/tree/main/skills/foo" → npx skills add https://github.com/owner/repo/tree/main/skills/foo --yes --agent claude-code

"Globally install foo/bar for all agents" → npx skills add foo/bar --yes -g --all

User says: find, search, discover, look for, browse skills 1. If user gave a keyword, pass it directly. 2. Run: ```bash npx skills find {keyword} --yes ``` 3. Present results to the user. If they pick one, follow the install operation. User says: check for updates, are my skills up to date Run: ```bash npx skills check --yes ``` Report which skills have updates available. Offer to update if any are found. User says: update skills, upgrade skills Run: ```bash npx skills update --yes ``` Report what was updated. User says: list skills, show installed skills, what skills do I have Run: ```bash npx skills add --list --yes ``` - Always use `--yes` to skip confirmation prompts. - Default to project scope unless the user explicitly says global. - Default to the current agent type only. Add others only if the user asks. - If a command fails, show the error output and suggest fixes (e.g., check the source URL, network). - After installing, confirm success and mention where the skill was installed.

<success_criteria>

  • The requested skill operation completed successfully.
  • Output was shown to the user confirming what happened.
  • Scope and agent targeting matched user intent (project/global, correct agent). </success_criteria>

Comments

Markdown supported