The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests untrusted data from external sources, specifically pull request descriptions and code diffs, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The agent is instructed to read pull request files and descriptions using gh pr view and gh pr diff, as well as repository-level documentation like README.md and CONTRIBUTING.md (SKILL.md).
Boundary markers: The instructions lack explicit boundary markers or "ignore embedded instructions" directives to distinguish between the SOP and the data being reviewed.
Capability inventory: The agent has the ability to post comments and reviews back to the repository using gh pr comment and gh pr review (SKILL.md).
Sanitization: The skill does not specify any sanitization or validation steps for the content retrieved from the pull requests.
[COMMAND_EXECUTION]: The skill instructs the agent to use standard GitHub CLI (gh) commands to interact with pull requests. While these are expected tools for the stated purpose, they provide the agent with the capability to perform repository-level operations and interact with the GitHub API.

strands-review