The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is instructed to read external content such as code changes, tickets, and existing plans. If these sources contain malicious prompts, they could influence the agent's behavior during the QA file generation or script execution phase. * Ingestion points: The skill instructions in SKILL.md direct the agent to read changes, tickets, plans, and existing QA documents. * Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are specified in SKILL.md. * Capability inventory: SKILL.md instructs the agent to write to .task//qa.md, create files in assets/, tests/, and scripts/ folders, and execute scripts. * Sanitization: There is no mention of sanitizing or validating input content in SKILL.md.
[COMMAND_EXECUTION]: The skill instructions in SKILL.md allow for the execution of scripts for database population. This capability, while intended for development environments, could be abused if the scripts are generated from untrusted ticket data.
[DYNAMIC_EXECUTION]: As described in SKILL.md, the agent generates support materials such as scripts and tests based on external task descriptions, involving the generation and potential execution of scriptable content.

qa-prepare