scan-community-repos

name	scan-community-repos
model	sonnet
description	Produces a compatibility catalog — skill-scan-report.md and skill-scan-catalog.json — classifying every SKILL.md in one or more GitHub repos or local paths as ready, normalizable, or incompatible, so you know exactly what normalization work lies ahead before committing to an import. Use when: "scan this repo for skills", "check which skills are dojo-compatible", "audit community skills", "catalog skills from a github repo", "find skill files in this repository", "assess community skill compatibility", "build a skill import report".
license	Complete terms in LICENSE.txt
category	skill-forge
triggers	["scan this repo for skills","check which skills are dojo-compatible","audit community skills","catalog skills from a github repo"]
tier	1
agents	["primary"]
tool_dependencies	["file_system"]
inputs	[{"name":"repo_paths","type":"string[]","description":"GitHub repo URLs or local paths to scan for SKILL.md files","required":true}]
outputs	[{"name":"scan_report","type":"ref","format":"cas-ref","description":"Compatibility catalog — skill-scan-report.md and skill-scan-catalog.json — classifying each SKILL.md as ready, normalizable, or incompatible"}]

Scan Community Repos

I. Workflow

Step 1: Accept and Validate Inputs

Accept input as:

• One or more GitHub repository URLs: https://github.com/owner/repo
• One or more local directory paths: /path/to/plugin-directory
• A mixed list of both

For each GitHub URL, confirm the repository is accessible. If the URL requires authentication or returns 404, log it as inaccessible and skip — do not abort the full scan.

For each local path, confirm the directory exists. If it does not, log and skip.

Output of this step: A validated input list with accessibility status for each source.

Step 2: Acquire Skill Files from Each Source

For GitHub URLs:

Clone each repository to a temporary directory:

git clone --depth=1 <repo-url> /tmp/skill-scan/<repo-name>

Use --depth=1 to avoid fetching full history. If git clone fails (network issue, private repo), log the error and skip the source.

For local paths:

Walk the directory tree in place. No copy needed.

Finding skill files:

Within each acquired source, find candidate skill files using two passes:

1. Exact match: files named exactly SKILL.md (case-sensitive)
2. Pattern match: any .md file that contains a YAML frontmatter block (---) and whose path contains skill in the directory name or filename

For pattern-match candidates, apply a quick filter: the frontmatter must contain at least one of name: or description: to be considered a skill candidate. Discard files that pass the filename heuristic but have no frontmatter at all (these are likely documentation, not skills).

Output of this step: A flat list of candidate file paths per source repo.

Step 3: Parse Each Candidate File

For each candidate file, parse:

1. Frontmatter extraction: Read lines between the first and second --- delimiters. Parse as YAML.
2. Body extraction: All content after the second ---.
3. Handle parse errors: If the YAML is malformed, classify the file as "incompatible" with reason "malformed frontmatter" and continue.