Claude Skill

skillify

Capture proven workflows as skills or guides, or move AGENTS.md procedures to their owning skill.

LLM Mart · 0 points · 2 views 0 listing impressions 0 install-command copies
Download fmind-dot-skills_skillify-3bbe830.zip · 13 KB

Install

skills CLI npx skills add https://github.com/fmind/dot/tree/main/skills/skillify
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install fmind-dot@llmmart
Git git clone https://github.com/fmind/dot.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole fmind/dot collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Skillify

Capture workflows as reusable skills or guides, extract AGENTS.md procedures, and route (Lesson Learned) session corrections to one canonical owner via lesson-learned. This skill owns workflow extraction, catalog conventions, and package rules.

Workflow

  1. Decide the owner first: for session corrections, decisions, or (Lesson Learned) proposals, run lesson-learned before touching a skill—prefer deterministic checks, tests, or code defaults over prose.
  2. Extract from the session: goal, trigger phrases, verified commands with flags, user corrections, decisions with reasons, dead ends, and required tools; strip session paths, one-off values, and secrets.
  3. Apply the admission rule in package rules before picking an owner or scope.
  4. Check neighbors: run rg -m1 '^description:' ~/.agents/skills/*/SKILL.md .agents/skills/*/SKILL.md (gh skill list scans host directories without descriptions) and inspect overlapping skills; extend an existing owner for the same workflow, create a new skill only for a distinct trigger, and link neighbors instead of copying.
  5. Choose scope:
    • Global (reusable, tool-generic): skills/<name>/ in fmind/dot (~/.agents/skills/<name>/), registered and checked via dot-skills (mise run check:skills + affected install tests).
    • Local (repo-specific commands or conventions): .agents/skills/<name>/ (read natively by agy; add .claude/skills -> ../.agents/skills only when Claude Code is used per agent-project; also registered via dot-skills inside fmind/dot).
  6. Author from skill.md: apply package rules; move long configs and examples to references/.
  7. Validate: frontmatter name matches directory, links resolve, resources are reachable, and mise run check:skills passes (or gh skill publish --dry-run .agents/skills in external repos).
  8. Exercise behavior: run the adoption check on substantial additions or routing edits (natural trigger, neighboring task, observable outcome in an isolated fixture); report unavailable host checks separately.
  9. Report: path, description, and scope.

Extracting from AGENTS.md

Keep personal defaults in global AGENTS.md and repo invariants in project AGENTS.md. Move procedures into their owning skill (new skill only for a distinct workflow), leave a routing cue, and sync via repository-docs.

Gotchas

  • Descriptions route, bodies instruct: state triggers in description, never a workflow summary.
  • Keep dates current: set created and updated on creation; bump updated on every edit.
  • Vendor external skills cleanly: follow vendor-skills instead of retyping upstream skills.
  • Follow portability when promoting a guide or publishing standalone.

Task guides

  • lesson-learned: Turn session corrections and decisions into minimal edits to checks, skills, AGENTS.md, or decision logs.

Documentation

Files (dot)
  • references
    • adoption-check.md 1.1 KB
      Loading…
    • catalog-review.md 1.5 KB
      Loading…
    • lesson-learned.md 2.4 KB
      Loading…
    • package-rules.md 7.6 KB
      Loading…
    • portability.md 1013 B
      Loading…
  • scripts
    • skill_loads.py 11 KB
      Loading…
  • templates
    • skill.md 957 B
      Loading…
  • SKILL.md 3.7 KB
    ---
    name: skillify
    description: "Route session lessons to checks, skills, AGENTS.md, or decision logs; author skills and guides."
    license: MIT
    metadata:
      kind: task
      author: Médéric HURIER (Fmind)
      source: github.com/fmind/dot/tree/main/skills/skillify
      created: "2026-09-02"
      updated: "2026-11-11"
    ---
    
    # Skillify
    
    Capture workflows as reusable skills or guides, extract `AGENTS.md` procedures, and route `(Lesson Learned)` session corrections to one canonical owner via [lesson-learned](references/lesson-learned.md). This skill owns workflow extraction, catalog conventions, and [package rules](references/package-rules.md).
    
    ## Workflow
    
    1. **Decide the owner first**: for session corrections, decisions, or `(Lesson Learned)` proposals, run [lesson-learned](references/lesson-learned.md) before touching a skill—prefer deterministic checks, tests, or code defaults over prose.
    1. **Extract from the session**: goal, trigger phrases, verified commands with flags, user corrections, decisions with reasons, dead ends, and required tools; strip session paths, one-off values, and secrets.
    1. **Apply the admission rule** in [package rules](references/package-rules.md#categories-and-scope) before picking an owner or scope.
    1. **Check neighbors**: run `rg -m1 '^description:' ~/.agents/skills/*/SKILL.md .agents/skills/*/SKILL.md` (`gh skill list` scans host directories without descriptions) and inspect overlapping skills; extend an existing owner for the same workflow, create a new skill only for a distinct trigger, and link neighbors instead of copying.
    1. **Choose scope**:
       - **Global** (reusable, tool-generic): `skills/<name>/` in `fmind/dot` (`~/.agents/skills/<name>/`), registered and checked via `dot-skills` (`mise run check:skills` + affected install tests).
       - **Local** (repo-specific commands or conventions): `.agents/skills/<name>/` (read natively by agy; add `.claude/skills -> ../.agents/skills` only when Claude Code is used per [agent-project](../agent-project/SKILL.md); also registered via `dot-skills` inside `fmind/dot`).
    1. **Author from [skill.md](templates/skill.md)**: apply [package rules](references/package-rules.md); move long configs and examples to `references/`.
    1. **Validate**: frontmatter `name` matches directory, links resolve, resources are reachable, and `mise run check:skills` passes (or `gh skill publish --dry-run .agents/skills` in external repos).
    1. **Exercise behavior**: run the [adoption check](references/adoption-check.md) on substantial additions or routing edits (natural trigger, neighboring task, observable outcome in an isolated fixture); report unavailable host checks separately.
    1. **Report**: path, description, and scope.
    
    ## Extracting from AGENTS.md
    
    Keep personal defaults in global `AGENTS.md` and repo invariants in project `AGENTS.md`. Move procedures into their owning skill (new skill only for a distinct workflow), leave a routing cue, and sync via [repository-docs](../repository-docs/SKILL.md).
    
    ## Gotchas
    
    - **Descriptions route, bodies instruct**: state triggers in `description`, never a workflow summary.
    - **Keep dates current**: set `created` and `updated` on creation; bump `updated` on every edit.
    - **Vendor external skills cleanly**: follow [vendor-skills](../agent-project/references/vendor-skills.md) instead of retyping upstream skills.
    - **Follow [portability](references/portability.md)** when promoting a guide or publishing standalone.
    
    ## Task guides
    
    <!-- guides:start -->
    
    - [lesson-learned](references/lesson-learned.md): Turn session corrections and decisions into minimal edits to checks, skills, AGENTS.md, or decision logs.
    
    <!-- guides:end -->
    
    ## Documentation
    
    - [Agent Skills specification](https://agentskills.io/specification) · [Catalog and usage review](references/catalog-review.md)
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.