The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill workflow includes instructions to run npm test and pytest on code provided in the pull request to calculate coverage. Executing code or tests from an untrusted source is a high-severity security risk, as it allows for arbitrary code execution within the agent's environment during the test run.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted content from pull request titles, bodies, and diffs.
Ingestion points: Fetches untrusted pull request data using gh pr view and gh pr diff commands.
Boundary markers: The instructions do not define delimiters or protective barriers to isolate the agent from embedded instructions in the ingested text.
Capability inventory: The agent has the ability to execute shell commands, perform network requests via curl, and run local test suites.
Sanitization: No input validation or sanitization is performed on the pull request content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill utilizes several CLI tools including gh, glab, and curl to perform its tasks. While these tools are used for their intended purpose, performing these operations using data derived from untrusted pull request content without strict validation creates a risk where command parameters could be influenced by an attacker.

pr-review-expert