Agent Skill Search
Portable skill catalog search for AI agents. MCP server + CLI.
Skills folders eat your agent's context window — every SKILL.md loads whether the task needs it or not. Move them into a catalog and the agent fetches a skill only when it's relevant. Use grep-like search terms to find the right one for your task.
Install
# recommended — isolated tool install
uv tool install agent-skill-search
# or
pip install agent-skill-search
The PyPI distribution is
agent-skill-search— the nameskill-searchon PyPI belongs to an unrelated package. The commands stayskill-searchandskill-search-mcp.
Usage
CLI
export SKILL_CATALOG_DIRS="$HOME/.hermes/skills"
skill-search search "react performance"
skill-search search "planning|prd|guide" # multiple keywords (OR search)
skill-search view "vercel-react-best-practices"
skill-search list
# Also scan well-known agent skills dirs (opencode, claude, codex, ...).
# Same-named skills are deduplicated; off by default.
skill-search --include-known list # or SKILL_INCLUDE_KNOWN=1
# Move skills from a source dir into the catalog (frees agent context).
# Note: `add` MOVES directories out of the source tree, it does not copy.
skill-search add ~/.hermes/skills
# Always shows the active dirs (stderr); inspect or change the default:
skill-search dirs # show default, config file, active dirs
skill-search dirs --set ~/skills # persist DIR as the default catalog
skill-search dirs --reset # back to ~/.agents/skills-catalog
MCP server
One command installs it into your client:
skill-search install claude-code # or opencode / codex / gemini / openclaw / hermes / pi / claude-desktop / chatgpt
See INSTALL.md for per-client details, manual config snippets, the ChatGPT remote-HTTP setup, and troubleshooting.
Agents without an MCP connection can use the CLI directly. Opt in to append a
short usage snippet to existing CLAUDE.md/AGENTS.md files (marker-checked,
append-only, never created from scratch):
skill-search install gemini --cli-hint
License
MIT — see LICENSE.
Release files for agent-skill-search 0.3.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| agent_skill_search-0.3.0.tar.gz | 27.3 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| agent_skill_search-0.3.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 48.7 kB
Release files / agent_skill_search-0.3.0.tar.gz
| Download URL | agent_skill_search-0.3.0.tar.gz |
|---|---|
| Size | 27.3 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
5e07e351a6a67ba903d3c7e212c2935d0540ca204feac79ea358418f16b10bcc
|
|
BLAKE2b-256 checksum How to use checksums |
fbd1e031a56c0b0dd690e7458b8ca175084a3c57e3f84469f9688fee5319b2eb
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Release files / agent_skill_search-0.3.0-py3-none-any.whl
| Download URL | agent_skill_search-0.3.0-py3-none-any.whl |
|---|---|
| Size | 21.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
4a1a3c365c100dfc7e1d397489f2495b52ee25cad9646324b1eb3ee71321bdb5
|
|
BLAKE2b-256 checksum How to use checksums |
6d8096ecdd09f1d4aa095171b2660afcafbc7609ae90300a661fbd6e44570cf0
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|