The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from task descriptions and external files which could be manipulated to influence agent behavior.\n
Ingestion points: The skill uses TaskGet to retrieve task descriptions and json:metadata blocks, and may also read from instances/<tag>/seed-briefing.md as specified in the SKILL.md file.\n
Boundary markers: There are no explicit instructions or delimiters defined to distinguish between instructional logic and the content being read from task metadata or briefing files, potentially allowing embedded instructions to be processed as valid configuration.\n
Capability inventory: The skill utilizes TaskUpdate to write back modified metadata (including a verifyCommand field intended for later execution by the executing-plans skill) and AskUserQuestion to facilitate user interaction.\n
Sanitization: The instructions do not mandate sanitization or validation of external data before it is interpolated into task metadata or shown to the user.

specifying-gates