The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data in the form of local code changes and new files, creating a surface for indirect prompt injection.
Ingestion points: Instructions in SKILL.md direct the agent to gather content via git --no-pager status --short, git --no-pager diff HEAD, and by reading the full content of new untracked files.
Boundary markers: The instructions do not specify the use of delimiters or clear separation between the agent's instructions and the untrusted code content being reviewed.
Capability inventory: The agent possesses capabilities to execute shell commands (git), read local files, and fetch external documentation via the context7 MCP tool.
Sanitization: There is no mention of sanitizing, escaping, or filtering the content of the diffs before they are processed by the LLM.
[COMMAND_EXECUTION]: The skill requires the execution of shell commands to interact with the local git repository.
Evidence: The skill explicitly instructs the agent to run git --no-pager status --short, git --no-pager diff HEAD, and git --no-pager diff --staged to gather context for the review. While these are standard development commands, they involve direct interaction with the host shell environment.

local-review