The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external platforms and test results which could potentially contain instructions aimed at influencing the agent's review decisions.
Ingestion points: Untrusted data is ingested into the agent context in SKILL.md via pull request descriptions (fetched using gh, glab, or twg in Step 0) and via visual/DOM diff data (fetched using meticulous agent in Steps 3 and 4).
Boundary markers: The skill does not explicitly instruct the agent to use delimiters or 'ignore embedded instructions' warnings when processing fetched external data.
Capability inventory: The skill has capabilities to execute shell commands, perform network requests to pull request platforms, and modify the state of Meticulous test runs (approving/rejecting diffs).
Sanitization: There is no evidence of sanitization or filtering applied to the external PR descriptions or DOM diff strings before they are presented to the agent.

meticulous-review