The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core workflow involves ingesting and analyzing README content from user-provided GitHub repositories to generate metadata and documentation.
Ingestion points: scripts/fetch_github_info.py retrieves README text, and scripts/import_github_skill.py downloads entire repository contents.
Boundary markers: In scripts/create_skill.py, external README content is placed within markdown blocks in the generated SKILL.md file.
Capability inventory: The skill utilizes subprocess.run for git operations, urllib.request for network access, and standard file I/O to create new skill structures.
Sanitization: The sanitize_name function in scripts/create_skill.py handles folder names, and basic regex cleaning is applied to README text, but there is no mechanism to filter embedded natural language instructions.
[EXTERNAL_DOWNLOADS]: The skill downloads repository metadata and source files from github.com and raw.githubusercontent.com. These operations are transparently documented as part of the skill's purpose to import external code.
[COMMAND_EXECUTION]: Multiple scripts (fetch_github_info.py, import_github_skill.py) use subprocess.run to execute git commands. These calls are implemented using argument lists and timeouts, which are standard security practices for preventing shell injection.

skill-factory