SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsadapt-skill

adapt-skill

Agents & models

View repository

Updated 10-16 05:03

About this skill

/adapt-skill — Port an external skill into your namespace

Sibling to create-skill. Different jobs:

  • create-skill → build a new skill from chat / dump / scratch
  • adapt-skill → take someone else's skill and make it yours

Step 1 — Fetch the source

Accept:

  • GitHub URL to a SKILL.md or a repo containing one (https://github.com/<owner>/<repo> or https://github.com/<owner>/<repo>/blob/main/SKILL.md)
  • agentskills.io URL or skills.sh URL
  • Local path to an existing skill on disk (other plugins installed in ~/.claude/plugins/)
  • Pasted SKILL.md content
# GitHub repo: clone shallow to a temp dir
git clone --depth 1 <url> /tmp/adapt-skill-<short-id>/
# OR fetch a single file
gh api -H "Accept: application/vnd.github.raw" repos/<owner>/<repo>/contents/SKILL.md > /tmp/adapt-source.md

Capture the commit SHA so attribution can point at a stable revision.

Step 2 — Analyze + classify

Read the source SKILL.md + any references/ it ships. Classify into three buckets per references/buckets.md:

Bucket What Example
Keep verbatim Format, schema, mechanic, scripts, frameworks The 38 decision questions, ffmpeg flags, a regex pattern, JSON schema for output
Adapt Tool defaults, paths, voice, naming, opinions Their video tool → your watch-video, their kanban → your pm, their voice → your voice, their "user" → "Corey" specifics
Add Cross-references to your existing skills, composition notes, your conventions "Composes with decide," "Saves to second-brain raw/ as <prefix>-," "Uses MLX-Whisper local"

Output the classification as a table for Corey to approve before writing anything. Don't silently rewrite — surface the changes.

Step 3 — License check

Read the source repo's LICENSE file. See references/license-check.md for the policy:

License Action
MIT / Apache-2.0 / BSD / ISC / CC0 ✅ Green — proceed
MPL-2.0 / LGPL 🟡 Yellow — adapt OK but warn about file-level reciprocity
GPL-2.0 / GPL-3.0 / AGPL ❌ Red — contagion risk for derivative work. Surface to Corey before proceeding.
Proprietary / no license ❌ Red — stop. No legal basis to copy.
Unclear 🟡 Yellow — ask Corey, default to skip if uncertain

For permissive licenses, attribution is the only requirement — handled in Step 6.

Step 4 — Pick target repo

Same as create-skill. Match by topic:

Repo What lives here
makerskills Personal cross-cutting operator work
cf-skills CF agency operations
marketingskills Generic marketing tactics (public)
nonfictionskills Writing a nonfiction book
fictionskills Writing fiction
youtubeskills Making YouTube videos

Infer if obvious; ask if ambiguous.

Step 5 — Rewrite the skill

For each bucket from Step 2:

Keep verbatim: copy as-is. Add a comment marker if helpful (<!-- from <source> -->).

Adapt:

  • Naming: rename to verb-noun if not (matches your convention: watch-video, read-book, create-skill, adapt-skill)
  • Tool swaps: their generic kanban tool → your pm; their video tool → watch-video; their transcript tool → watch-video transcript; their note system → second-brain; their decision framework → decide
  • Voice: apply your voice rules — direct, conviction-coded, no AI-slop. For social-adjacent skills, apply the link-placement rule.
  • Paths: their ~/outputs/ → your ~/Documents/<skill>-<...>/ convention; their config files → your references/ pattern
  • Names + context: their "the user" / generic examples → Corey + his portfolio context (CF, Swipe Files, Magister, Truelist, etc.) where the skill needs it

Add:

  • Cross-references to existing makerskills/cf-skills/marketingskills siblings
  • Composition notes (which other skills this calls or feeds)
  • Your conventions (frontmatter version, references/ subdir structure, BACKLOG entry if it spawns sub-ideas)

Step 6 — Attribution file

Write references/attribution.md:

# Attribution

This skill was adapted from an external source. Original credit:

- **Source**: <original SKILL.md URL>
- **Repository**: <repo URL>
- **Author**: <author name + handle if known>
- **Commit SHA**: <SHA at time of adapt>
- **License**: <license name + URL>
- **Adapted**: <YYYY-MM-DD>
- **Adapted by**: Corey Haines

## What was kept verbatim
- <piece>
- <piece>

## What was adapted
- <piece> → <new version>
- <piece> → <new version>

## What was added
- <piece>
- <piece>

## License compliance

<copy the upstream LICENSE text here if MIT/BSD/Apache requires it, OR reference where in this repo it lives>

## Upgrade path

To check for upstream changes:
\`\`\`bash
git ls-remote <repo-url> HEAD
\`\`\`

Compare returned SHA to the "Commit SHA" above. If different, run `/adapt-skill` again on the same source to re-port (you'll get a 3-way merge view: upstream changes vs your local adaptations vs current).

For permissive licenses requiring notice (MIT, Apache, BSD), the LICENSE text either gets copied into this file or the source LICENSE file gets copied to the skill dir.

Step 7 — Scaffold + commit

Same as create-skill Step 6–7:

  • Write SKILL.md + references/ (including attribution.md)
  • Append to target repo's README skill table
  • Commit with message: "Adapt <name> skill from <source-name>"
  • Push

Step 8 — Report + offer follow-ups

  • Commit hash + new skill path
  • Show the three-bucket classification one more time (what was kept / adapted / added) so Corey sees the diff clearly
  • Offer:
    • "Want to flesh out the adaptations? Some defaults may still need your touch."
    • "Want me to run compound-engineering:heal-skill for a QA pass?"
    • "Should I set up an upstream-check reminder via loop or compound-engineering:schedule?"

Composes with

  • create-skill — sibling. Use that for from-scratch / from-chat / from-dump. Use this for from-external.
  • compound-engineering:create-agent-skill — call into for format expertise during the rewrite
  • compound-engineering:heal-skill — QA pass after adaptation
  • compound-engineering:skill-creator — deep best-practices reference
  • loop / compound-engineering:schedule — for periodic upstream-check (future enhancement)

Examples of adaptations already done informally in this repo

These were done by hand during the conversation that built this plugin. Future ones should run through /adapt-skill:

Source Adapted into What was taken
zarazhangrui/frontend-slides slide-deck "Show, don't tell" pattern, density modes, anti-AI-slop rules, Playwright export approach
mvanhorn/last30days-skill deep-research (composes) Used as a sub-tool, not adapted directly
Hermes' llm-wiki second-brain (reference) Schema validation — referenced but not ported
garrytan/gbrain second-brain People-as-entities, scheduled maintenance pattern
garrytan/gstack Naming inspiration Short brand-able skill repo names

These already exist with informal attribution in their SKILL.md bodies. If you want, run /adapt-skill retroactively on any of them to get a proper references/attribution.md file.

Notes on quality

  • Surface the bucket classification before writing. Don't silently rewrite — Corey approves what changes and what stays.
  • Attribution is non-negotiable. Every adapted skill ships with references/attribution.md. Credit + license compliance + future upgrade path live here.
  • License is the gate. Don't proceed on GPL/proprietary without explicit Corey approval. Defaults to stop, not warn.
  • Don't import dependencies you don't need. If the source ships 5 reference files and you only need 2, leave the other 3. Carrying weight = future maintenance.
  • Tool-swap aggressively. If your stack has a better-fit tool than the source's choice, swap. The adaptation IS the value — verbatim copies aren't.