The Agent Skills Directory

Indirect Prompt Injection Surface: The skill processes untrusted pull request diffs which could potentially influence the agent's review verdict or actions.
Ingestion points: Pull request diffs are fetched and processed in Step 5 using the gh pr diff command.
Boundary markers: The skill does not currently use specific delimiters or instructions to differentiate between untrusted diff content and its own internal instructions.
Capability inventory: The skill has the capability to post comments via the GitHub API and autonomously approve or request changes on pull requests.
Sanitization: Diff content is reviewed for specific patterns like secrets or breaking changes, but there is no general sanitization of the input text before processing.
External Data Verification: The skill fetches metadata from the NuGet API to verify the existence of package versions. This data is processed using a local script, which is a standard procedure for confirming deployment readiness.
Command Execution: The skill utilizes the GitHub CLI (gh) and standard system tools like grep and node to perform its tasks. These tools are used appropriately within the context of managing repository workflows.

review