FAQ

FAQ

Home | Quick Start | Safety and Data

Can I install it from npm?

Not yet. The npm package name is reserved, but the documented installation path is git clone:

git clone https://github.com/GrubbyLee/skill-manager.git
cd skill-manager
node scripts/install.mjs

What does the bare skm command do?

skm is the same as skm status. It prints a health overview: total skills, MCP count, never-used skills, duplicate installs, session log size, health score, and suggested next commands.

Does skm scan modify my Claude Code or Codex setup?

No. It writes ~/.skill-manager/catalog.json for skm's own inventory cache. It does not edit skills, configs, MCP servers, or session logs.

How does skm audit decide a skill was used?

It looks for real usage signals in session logs. Codex may inject all skill paths into context, but skm only counts actual skill reads or calls as usage.

Why are Codex-only MCP servers marked unobservable?

The current usage signal for idle MCP comes from Claude-side logs. Codex-only MCP servers are reported as unobservable instead of being treated as idle.

How do I export every scanned skill or MCP?

Use JSON output:

skm scan --json > scan.json
skm list --json > skills.json
skm list --mcp --json > mcp.json

How do I make the graph less crowded?

In the HTML graph:

• turn off same category and shared platform first
• enable important-node mode
• hide never-used skills
• use search to focus a topic
• drag local clusters by hand
• click fit view after filtering

What should I run before cleaning anything?

skm doctor
skm scan
skm
skm risks
skm report --format html --output skm-report.html
skm sessions --clean --days 30 --keep 3 --dry-run

Only remove --dry-run after reviewing the plan.