All skills
aktsmm avatar

/skill-finder

@4461e59
by yamapanaktsmm/agent-skills26 stars
4

Search, install, and manage Agent Skills locally and from GitHub, then help decide whether the task really needs a skill or another customization primitive. Use when looking for skills, installing skills, managing a skill collection, or choosing between a skill, prompt, instruction, or agent.

Use this Skill: https://skilld.dev/gh/aktsmm/agent-skills/skill-finder

This session only. Nothing lands on disk.

referencescustomization-routing.md

≈715 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Customization Routing

Use this guide before recommending a skill.

Decide Whether the User Needs a Skill

User Intent Best Fit
"いつもこのルールで動いて" Instruction
"この slash task を作りたい" Prompt or Skill
"専用 persona で動かしたい" Agent
"危険コマンドを止めたい" Hook
"再利用できる workflow を配布したい" Skill

Fast Rules

  • If the user asks for a reusable package with references, scripts, or templates, recommend a skill.
  • If the user only needs a focused slash command, recommend a prompt first.
  • If the user mainly needs role boundaries or tool restrictions, recommend an agent.
  • If the user wants behavior enforced automatically, recommend a hook.

Response Pattern

When the ask is not best served by a skill, say so before searching.

This sounds closer to an instruction than a skill because the behavior should apply automatically.

Do you want:

1. An instruction template?
2. Skills anyway?

If the user still wants skills, continue the skill search normally.

Recommendation Heuristics

When multiple skills match, prefer:

  1. Official or curated source
  2. Clear description with trigger conditions
  3. Good resource structure (scripts/, references/, assets/)
  4. Recently maintained index entries

If a curated native skill and an upstream / mirrored skill cover the same domain, prefer the curated native skill as the recommendation default. Treat the mirrored skill as a reference or intake source unless the user explicitly asks for that upstream workflow.

Collection Management

When deciding whether a discovered skill belongs in a shared collection, classify it first:

State Meaning Action
Candidate Interesting but unproven Keep in a sandbox/source list; do not present as default
Adopted Used successfully on real tasks Star it or install locally for repeat use
Shared Useful across people/repos Add to the maintained index or private marketplace source
Retire Duplicate, stale, or under-triggering Remove from recommendations or replace with a better source

Promotion should be based on observed usage, clear trigger fit, and maintainability rather than novelty.

Source inspiration: Anthropic, "Lessons from building Claude Code: how we use skills" - https://claude.com/blog/lessons-from-building-claude-code-how-we-use-skills

Source: SKILL.md on GitHub

3 warnings15d5 checks · Risk SAFE
  • Gen Agent Trust Hub15d

    This skill acts as a manager for agent skills, allowing users to search, install, and update them from various GitHub repositories. While it includes safety features like requiring user approval for unknown hosts and displaying requirements before installation, it facilitates the download of external, unverified code from community sources and instructs the agent to execute commands silently, which reduces visibility into automated operations.

  • Socket15d

    2 alerts: gptSecurity, gptAnomaly

  • Snyk15d

    Risk: MEDIUM · 1 issue

  • Runlayer7mo

    12/12 files flagged

  • ZeroLeaks5mo

    Score: 93/100 · 2 sections analyzed

Signed by skilld at 4461e59. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 18 hours ago.

Activeupdated last month
argument-hint
探したい skill の用途やキーワード
user-invocable
true
metadata
{
  "author": "yamapan (https://github.com/aktsmm)"
}

README badge

README badge for aktsmm/agent-skills/skill-finder