The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill requires the agent to verify pull requests by checking out the code and running "relevant tests or commands" (Step 3 in SKILL.md). Since pull requests can originate from external, untrusted contributors, this instruction leads the agent to execute arbitrary code from an unknown source.
[DYNAMIC_EXECUTION]: The verification process involves the execution of tests or commands based on the PR diff, which constitutes dynamic execution of untrusted code at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from issue and PR bodies, comments, and diffs (Step 1 in SKILL.md). This data is then used to guide subsequent actions, including the execution of commands, without adequate isolation.
Ingestion points: Issue tracker (body, comments, author), Pull Request (body, comments, author, diff).
Boundary markers: None identified for input data; a disclaimer is added to output but not to input processing.
Capability inventory: Checking out code, running shell commands, running tests, posting comments to the issue tracker, and writing to the filesystem (.out-of-scope/ directory).
Sanitization: No evidence of sanitization, validation, or filtering of the external content before it is processed or used to drive execution.
[COMMAND_EXECUTION]: The skill invokes other commands and skills, such as /grilling, /domain-modeling, and /setup-leandrocfe-skills, based on interpretations of natural language requests from potentially external sources.

triage