The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process code diffs and pull request metadata from external, untrusted sources. This data could contain embedded malicious instructions intended to manipulate the agent's review process or output.
Ingestion points: The pr-review.md workflow fetches repository content via git clone, and SKILL.md processes output from git diff and git show.
Capability inventory: The skill has access to shell execution for version control and file system operations within the /tmp directory.
Sanitization: No specific input sanitization is defined for the content of the code diffs or PR descriptions.
Boundary markers: The skill does not utilize explicit delimiters or warnings to separate untrusted code from its own operational instructions.
[COMMAND_EXECUTION]: The skill utilizes shell commands to interact with the local environment and manage temporary review data.
Evidence: The allowed-tools list in SKILL.md includes git (diff, log, show, etc.), gh api, and file system commands like rm -rf /tmp/vibe-review-*, ls, and mkdir.
[EXTERNAL_DOWNLOADS]: The skill performs automated downloads of external source code based on user-provided repository URLs.
Evidence: The pr-review.md file defines a multi-step process for cloning repositories from platforms including GitHub, GitLab, and Gitee.
[DYNAMIC_CONTEXT_INJECTION]: The skill uses dynamic context injection to determine the current repository and working directory during initialization.
Evidence: SKILL.md executes git remote -v and pwd using the !command syntax to establish environment context.

vibe-review