The Agent Skills Directory

Command Execution: The skill uses shell commands like git status, git diff, pytest, and pre-commit to perform its review and testing tasks. This is a core capability that allows the agent to interact with the local file system and execution environment.
Indirect Prompt Injection: The skill analyzes uncommitted code changes, which could potentially contain embedded instructions intended to manipulate the agent's review.
Ingestion points: Data enters the context via the git diff workflow in SKILL.md.
Boundary markers: There are no specified delimiters used to separate the ingested code from the agent's instructions.
Capability inventory: The agent has the ability to run shell commands and modify local files.
Sanitization: The skill does not explicitly sanitize the contents of the code diffs before processing.

adk-review