The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to verify external pull requests by checking them out and running 'relevant tests or commands'. Because this code originates from untrusted external contributors, it provides a direct path for executing malicious code on the host system during the triage process.
[COMMAND_EXECUTION]: The agent is prompted to execute shell commands and test suites to verify the functionality of pull requests and explore the codebase for redundancy.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from issue trackers and PR diffs to drive its triage logic, creating a vulnerability surface where malicious content could influence the agent's behavior.
Ingestion points: Issue and PR bodies, comments, and diffs (referenced in SKILL.md).
Boundary markers: None identified for delimiting external content from agent instructions.
Capability inventory: The skill has the ability to read and write files (specifically in the .out-of-scope/ directory), execute arbitrary system commands (tests/setup), and invoke other skills like /grilling.
Sanitization: No sanitization or validation mechanisms are mentioned for the processed external content.

triage