The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection, as it synthesizes information from potentially untrusted sources like repository code comments and chat history to generate its final output. Use of this data to write to an issue tracker provides a path for embedded malicious instructions to influence the repository's project management state. \n- Ingestion points: Repository files and conversation context (SKILL.md). \n- Boundary markers: The instructions do not define delimiters or explicit warnings to ignore embedded commands within the ingested data. \n- Capability inventory: The skill performs file system read operations to explore the repo and uses a tool to publish content to an external issue tracker. \n- Sanitization: There is no mention of sanitizing or validating the generated content before publication, although the skill does instruct the agent to check proposed modules with the user first.

to-prd