The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and synthesize information from potentially untrusted sources such as issue tickets, pull requests, and repository diffs into an execution context that guides downstream autonomous agents.
Ingestion points: External data enters the context via the 'rediscovered repository, ticket, PR, or diff facts' (SKILL.md, Step 1).
Boundary markers: The skill uses a 'Locked in (tell me if this is wrong)' announcement to establish a shared understanding and verify conclusions with the user before proceeding, acting as a manual validation gate (SKILL.md, Step 6).
Capability inventory: Settled decisions and the execution context are passed to highly capable downstream skills including /analyze, /task-with-tests, and /task (SKILL.md, 'If this is a user one-off' section).
Sanitization: There are no explicit technical sanitization or filtering steps for the ingested data; the skill relies on the agent's reasoning capabilities and user oversight of the final locked-in message to prevent malicious instructions from being adopted.

grill-me