The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions in references/readme.md require the agent to fetch external content from https://skills.sh/docs and use it to extract formatting rules for README files. This creates a surface for indirect prompt injection where malicious instructions could be embedded in the remote documentation.
Ingestion points: The <required_reading> block in references/readme.md triggers a fetch to https://skills.sh/docs.
Boundary markers: Absent. The skill lacks explicit instructions or delimiters to prevent the agent from obeying instructions that might be contained within the fetched content.
Capability inventory: The skill is capable of scanning the local repository using git or ripgrep and writing markdown files. It does not possess direct shell execution or network exfiltration tools beyond the initial fetch.
Sanitization: Absent. The skill does not validate or sanitize the fetched content before using it to generate documentation.
[EXTERNAL_DOWNLOADS]: The skill fetches documentation from https://skills.sh/docs to verify installation commands and badge formats. This is a functional requirement for the skill to provide accurate platform-specific instructions.

write-docs