The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs project state inspection using git status and validates configuration syntax through standard local utilities, specifically ruby -e 'require "yaml"; YAML.load_file(...)' and python3 -m json.tool. These operations are localized to the repository and are used for verification purposes.
[EXTERNAL_DOWNLOADS]: References official documentation for CodeRabbit (docs.coderabbit.ai) and Greptile (greptile.com). These are well-known services relevant to the skill's primary function and do not include instructions for automated remote code execution or untrusted downloads.
[PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection by reading project-specific context such as style guides, agent instructions, and existing configuration files to tailor the reviewer setup. While these ingestion points could potentially contain malicious instructions, the skill uses the data solely for local configuration management within the repository's scope.
Ingestion points: Reads greptile.json, .coderabbit.yaml, project style guides, and architecture documentation (SKILL.md Step 1, 2, 5).
Boundary markers: No explicit instructions are provided to the agent to ignore embedded instructions in the processed files.
Capability inventory: File read/write access, git status commands, and local syntax validation using Python and Ruby (providers/coderabbit.md, providers/greptile.md).
Sanitization: No specific sanitization or filtering of the ingested project data is performed.

ai-reviewer-setup