The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/review_token.py executes the local openssl binary to sign JSON Web Tokens (JWTs) for GitHub App authentication.
Evidence: The command is constructed as a list ["openssl", "dgst", "-sha256", "-sign", str(key)] and executed via subprocess.run in scripts/review_token.py (Line 134). This approach avoids shell interpretation and injection vulnerabilities.
Evidence: The script also generates shell commands intended for consumption by other skills, using shlex.quote to ensure the generated paths and arguments are safely escaped for execution in a shell context (Line 245).
[INDIRECT_PROMPT_INJECTION]: The skill provides an interface to resolve identity information from a repository's docs/agents/issue-tracker.md file, which constitutes an ingestion point for untrusted data.
Ingestion points: The Resolve and Setup instructions in SKILL.md describe reading parameters like Reviewing app and Client id from the tracker document within a repository.
Capability inventory: The skill possesses the ability to execute local commands (openssl via subprocess.run) and perform network operations via urllib.request to communicate with the GitHub API.
Boundary markers: The SKILL.md instructions specify looking for a specific block structure (## Reviewing identity) to reduce the risk of processing irrelevant content.
Sanitization: All values parsed from the external tracker document that are subsequently used in generated commands are passed through shlex.quote in scripts/review_token.py to prevent command injection.

review-bot