The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run local Python scripts (scripts/init_skill.py and scripts/package_skill.py) to automate the scaffolding and distribution of new documentation skills. These scripts use standard Python libraries (pathlib, zipfile, sys) to perform file system operations and do not initiate any network connections.
[SAFE]: The validation script (scripts/quick_validate.py) uses yaml.safe_load to process skill metadata, preventing arbitrary code execution during the parsing of configuration files.
[SAFE]: Indirect Prompt Injection Risk: The workflow involves an 'Explorer agent' that fetches and processes documentation from external URLs and GitHub repositories to identify best practices. This ingestion surface is well-mitigated:
- Ingestion points: External documentation sources identified by the user in Step 3.
- Boundary markers: Although the provided prompts do not include explicit delimiter warnings, they are focused on specific technical extraction tasks.
- Capability inventory: The skill's capabilities are limited to local file system writes based on templates.
- Sanitization: Step 4 of the workflow requires human curation of findings against specific criteria before they are incorporated into the final skill, acting as a critical security checkpoint.

dev-skill-creator