The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains a surface for indirect prompt injection because it ingests untrusted data from external sources such as repository documentation (e.g., AGENTS.md, REPO_STYLE.md) and git diff outputs (SKILL.md, Workflow steps 1 and 3). There are no specified boundary markers or sanitization procedures to filter instructions embedded in this data. However, the security impact is minimal as the agent's capabilities are limited to observational reporting and the use of restricted version control tools.\n- [COMMAND_EXECUTION]: The skill instructions specify the use of version control commands, including git diff and git status, to inspect the code being reviewed (SKILL.md, Workflow step 3). These commands are static and do not incorporate unvalidated user input, making them safe within the context of the skill's intended purpose as a developer tool.

old-python-code-review