agentsclimarketplace

Skm doctor agent skills

Skill weiox/skm/skills/skm-doctor-agent-skills

Agent Skill Manager — discover, install, link, and upgrade skills for Claude Code & Codex.

Install
npx -y skills add weiox/skm --skill skm-doctor-agent-skills

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when agent skill entrypoints for Codex and Claude Code may be broken, duplicated, unmanaged, or drifting away from `~/.skm`, and you want a structured diagnosis before fixing anything.

SKILL.md

3.4 KB, as published. Nobody here has run it

Doctor Agent Skills

Overview

Before reorganizing or reinstalling skills, inspect the current state first.

This skill diagnoses the two entrypoint layers:

  • ~/.agents/skills
  • ~/.claude/skills

and checks whether they are still aligned with the managed source tree under ~/.skm.

When to Use

Use this skill when:

  • a skill is visible in one agent but not the other
  • old symlinks keep showing up after reorganizing
  • you suspect broken links or unmanaged links
  • you want a quick health check before running larger migrations
  • you just created a new skill under ~/.skm/personal and it still does not show up in Codex or Claude Code

Do not use this skill for project-specific repository guidance or for installing new skills from links.

Core Rule

Diagnose first, then fix.

Run the doctor script before making cleanup changes:

bash ~/.skm/skills/skm-doctor-agent-skills/scripts/skm-doctor-agent-skills.sh

What It Reports

The script classifies each entrypoint symlink as:

  • OK — managed and resolves inside skm
  • BROKEN — symlink exists but target does not
  • UNMANAGED — symlink points outside the managed skm tree
  • CONFLICT — more than one declared skill source shares the same skill name

Common Visibility Pitfalls

  • In this setup, Codex runtime discovery comes from ~/.agents/skills, not from ~/.codex/skills.
  • Creating a skill under ~/.skm/personal is not enough by itself; the export and entrypoint layers still need to be rebuilt.
  • A clean doctor result only says the runtime links are structurally healthy. It does not refresh the skill list inside an already-running Codex session.

Workflow

1. Run the doctor

bash ~/.skm/skills/skm-doctor-agent-skills/scripts/skm-doctor-agent-skills.sh

2. Read the categories

  • If you see BROKEN, rebuild or remove stale entrypoints
  • If you see UNMANAGED, decide whether the entry should move into skm
  • If you see CONFLICT, stop and resolve the duplicate source-of-truth problem before running sync
  • If everything is OK, the entrypoint layer is at least structurally healthy
  • If the structure is healthy but a newly created skill is still missing, run skm-sync-agent-skills and then start a new Codex session

3. Apply the next skill

After diagnosis:

  • use skm-organize-agent-skills when source-of-truth layout is wrong
  • use skm-install-linked-agent-skills when a vendor package is missing
  • use bootstrap.sh --force when the source is correct and only the entrypoints need rebuilding

Common Mistakes

  • Fixing links blindly before inspecting current state
  • Treating any symlink as healthy just because it exists
  • Leaving unmanaged links in ~/.agents/skills or ~/.claude/skills
  • Ignoring duplicate skill names across skills/, personal/, and vendor/
  • Assuming Codex and Claude Code should always expose the same shape of entrypoints

Completion Checklist

Before calling the diagnosis complete:

  1. the doctor script has run on the current machine
  2. broken, unmanaged, conflicting, and healthy states are clearly distinguished
  3. the next remediation step is chosen based on the diagnosis output

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.