The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/build_social_catalog_facts.py utilizes subprocess.run to execute local git commands. This functionality is used to programmatically retrieve repository metadata, such as origin URLs and commit hashes, for manifest records. The command arguments are restricted to internal metadata queries and do not process unsanitized user input.
[PROMPT_INJECTION]: The skill processes untrusted external data (article text, repository READMEs, and social media posts) to generate image prompts, creating a surface for indirect prompt injection. The skill mitigates this risk by implementing a strict 'source grounding' contract, using structured prompt blocks to isolate external facts from instructional logic, and requiring manual or automated validation of the final image output.

soia-media-generate-article-image