agentsclimarketplace

Skillify

Skill leadgenjay/Reeper/plugins/reeper/skills/skillify

Safely adapt external repositories into an existing codebase through evidence-first analysis, conflict interviews, and verified integration.

Install
npx -y skills add leadgenjay/Reeper --skill skillify

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 13 days oldThe repository was created 13 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Turn a repository or a stable workflow inside a repository into a callable Agent Skill or Claude Code plugin. Use when the user wants a repo exposed as a slash command, reusable skill, plugin, subagent, or installable marketplace package. Performs workflow discovery, a focused interface interview, progressive-disclosure design, scaffolding, validation, and test prompts. Invoke explicitly as /reeper:skillify.

SKILL.md

3.7 KB, 699 tokens by cl100k_base, as published. Nobody here has run it

Reeper Skillify

Package a stable workflow interface, not a repository dump.

Read:

  • ${CLAUDE_SKILL_DIR}/../../references/packaging.md
  • ${CLAUDE_SKILL_DIR}/../../references/security.md
  • ${CLAUDE_SKILL_DIR}/../../references/artifact-contract.md

1. Analyze before designing

Identify:

  • what users repeatedly ask the repository to do
  • deterministic operations that belong in scripts
  • judgment-heavy operations that belong in instructions
  • inputs, outputs, side effects, permissions, and failure modes
  • required CLIs, services, MCP servers, environment variables, and credentials
  • whether it is primarily reference knowledge, an action workflow, a specialized agent, an MCP integration, or a standalone application

Use Repomix or the fingerprint script for large repositories, but read the actual entry points and docs before deciding on commands.

2. Focused interface interview

Ask exactly one material question at a time. Resolve:

  • the primary callable job
  • expected arguments
  • whether Claude or only the user may invoke it
  • read-only versus write/destructive behavior
  • permission boundaries and approval gates
  • required integrations and fallback behavior
  • persistent state and generated artifacts
  • installation scope: project skill, personal skill, plugin, or marketplace
  • command names and discoverability triggers
  • success and failure outputs

Recommend the narrowest interface that satisfies the job.

3. Design

Prefer this progression:

  1. one action skill
  2. supporting references and deterministic scripts
  3. subagents only when context isolation or specialized permissions materially help
  4. plugin wrapper when distribution or multiple components require it
  5. marketplace wrapper only when installation from a Git repository is desired

Keep each SKILL.md under 500 lines when practical. Put large references in references/, executable helpers in scripts/, templates/assets in their own directories, and eval prompts in evals/.

4. Scaffold

For a standalone skill:

python3 "${CLAUDE_SKILL_DIR}/../../scripts/scaffold_skill.py" \
  --output <output-dir> \
  --name <skill-name> \
  --description "<what and when>"

For a plugin, create:

<plugin>/
├── .claude-plugin/plugin.json
├── skills/<skill-name>/SKILL.md
├── agents/                 # only when justified
├── scripts/
├── references/
└── evals/

For marketplace distribution, add a repository-root .claude-plugin/marketplace.json and keep the plugin self-contained beneath its declared source path.

5. Validate and test

Create at least:

  • three should-trigger prompts
  • three should-not-trigger prompts
  • two happy-path execution prompts
  • two ambiguity/conflict prompts
  • one denied or unsafe operation prompt
  • one missing-dependency prompt

Validate locally when Claude Code supports it:

claude plugin validate <marketplace-or-plugin-path>

Test the exact installation and invocation commands. Report any validation that could not be run.

6. Deliver

Provide:

  • generated path
  • installation command
  • callable command(s)
  • required dependencies/integrations
  • permission and security notes
  • test/eval status
  • version bump guidance

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.