The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts and standardized command-line tools to gather repository context.
The file SKILL.md invokes a local script scripts/collect-review-context.sh to automate the discovery of diffs and relevant local documentation.
Multiple files, including references/git-context.md, utilize git, gh (GitHub CLI), and rg (ripgrep) to query metadata, branch history, and file contents.
These operations are used solely for read-only context gathering and are consistent with the tool's primary function.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository environment, which presents a surface for indirect prompt injection.
Ingestion points: As described in SKILL.md and scripts/collect-review-context.sh, the tool ingests code diffs, pull request titles/bodies, commit messages, and project-specific documentation like AGENTS.md and README.md.
Boundary markers: The skill relies on standard markdown headers and lacks explicit instructions or delimiters to the model to ignore instructions potentially embedded within the code or documentation files being reviewed.
Capability inventory: The agent has access to git (including fetch operations), gh, and rg, along with file system read access and the ability to spawn subagents.
Sanitization: The skill does not employ specific sanitization or escaping techniques for the external data it reads into the context before processing.

singular-code-review