Skip to main content

Test

skillgrep

Search local Agent Skill collections without loading them all into every agent session.

Agents normally advertise every enabled skill's name and description, then load the selected SKILL.md on demand. skillgrep keeps infrequent private or team collections outside that catalog. One routing skill searches their metadata and resolves only the chosen file.

It does not clone, update, install, or execute registered skills.

Install

$ npx skills add vrslev/skillgrep

The skill invokes the dependency-free CLI with uvx skillgrep.

Configure

$ uvx skillgrep add ~/code/team-skills --name team
added  team  2

The name becomes the stable namespace in results such as team:release-workflow. It defaults to the directory name.

Configuration follows the XDG application convention:

${XDG_CONFIG_HOME:-~/.config}/skillgrep/config.json

There is no standard Agent Skills configuration directory. Normal setup uses add, rm, and ls; manual configuration is optional:

{
  "version": 1,
  "registries": {
    "personal": "~/code/personal-skills",
    "team": "~/code/team-skills"
  }
}

The file is created with mode 600 on POSIX systems. Set SKILLGREP_CONFIG or pass --config to use another location.

Example

$ uvx skillgrep q "release incident"
team:incident-release-check  Verify release state and collect incident evidence before taking action.
personal:release-notes  Draft concise release notes from a local Git history.

$ uvx skillgrep path team:incident-release-check personal:release-notes
/Users/example/code/team-skills/delivery/incident-release-check/SKILL.md
/Users/example/code/personal-skills/writing/release-notes/SKILL.md

q returns at most three one-line matches by default; --top N changes the limit. It omits paths, remotes, and configuration locations. path reveals only the selected files. Ranking is deterministic and uses the query, skill name, and description—no embeddings or model calls.

Commands

skillgrep add PATH [--name NAME]  Register a local collection
skillgrep rm NAME                 Remove a registry
skillgrep ls [--paths]            List registries and skill counts
skillgrep q QUERY [--top N]       Query names and descriptions
skillgrep path SKILL [SKILL ...]  Print selected SKILL.md paths

Registry-qualified identifiers prevent collisions between collections. Duplicate names inside one registry are errors.

Privacy and trust

  • Runtime commands read only explicitly registered local directories and make no network requests. Installation and uvx package retrieval require network access.
  • Registry paths stay in the local XDG config rather than agent settings or a public dotfiles repository.
  • Descriptions are search metadata; do not put secrets in skill frontmatter.
  • Registered collections must be trusted because selected files become agent instructions.
  • Symlinked skill files resolving outside their registered root are ignored.

How this differs from native discovery

The comparison reflects documented behavior available on 2026-08-26.

System Initial discovery Detailed instructions
Codex Enabled skill metadata is placed in the model-visible catalog. The selected SKILL.md is read on demand.
Claude Code Metadata for model-invokable skills is available for routing; skills may also be manual-only. Skill content loads when invoked.
OpenCode Permitted names and descriptions appear in the native skill tool. The skill tool loads selected content.
Gemini CLI Names and descriptions of enabled skills enter the system prompt. activate_skill loads selected content after consent.
skillgrep Only its routing skill is installed. Registered collections remain outside native discovery. The agent queries locally and resolves one chosen file.

Native discovery remains preferable for a small, frequently used set. skillgrep covers the larger searchable tail. It is agent-mediated discovery, not a patch to an agent's native prompt assembler.

Metadata

Release files for skillgrep 0.2.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for skillgrep 0.2.1
File Size Uploaded
skillgrep-0.2.1.tar.gz 8.5 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for skillgrep 0.2.1
File Interpreter ABI Platform
skillgrep-0.2.1-py3-none-any.whl Python 3 none any Details

Total release size: 18.4 kB

Release files / skillgrep-0.2.1.tar.gz

Download URL skillgrep-0.2.1.tar.gz
Size 8.5 kB
Tags Source
SHA-256 checksum
How to use checksums
bbcb6398b02701b38cd107109d09abc8edca44a3e4a32fde4bde1077eaf0c51e
BLAKE2b-256 checksum
How to use checksums
52a886f068fef4e5e09cfc0a59f221c45c1ba6c549f7088f5a6a2bee01b672c6
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.12.6 {"installer":{"name":"uv","version":"0.12.6","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Release files / skillgrep-0.2.1-py3-none-any.whl

Download URL skillgrep-0.2.1-py3-none-any.whl
Size 9.9 kB
Tags Python 3
SHA-256 checksum
How to use checksums
aeb949682d2d456c5954665e12ff63156275902828194d4370537159a2246121
BLAKE2b-256 checksum
How to use checksums
1bfde1ffdef1d4fdbee941f52c6a7c3c720115ac8f3b20f79a32dbb34484b552
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.12.6 {"installer":{"name":"uv","version":"0.12.6","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Release history Release notifications | RSS feed

0.2.2

2 release files

This release

0.2.1 This release

2 release files

0.2.0

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page