The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from a git diff and passes it to subagents, which could lead to instructions embedded in the code diff being executed by the subagents.
Ingestion points: The content generated by git diff origin/main is captured and interpolated into the prompt for subagents in the 'Launch' section of SKILL.md.
Boundary markers: The prompt template lacks explicit delimiters or structural markers to wrap the [INSERT FULL DIFF] variable, which may allow content in the diff to blend with the agent's instructions.
Capability inventory: The skill launches independent subagents (sub-processes) and provides them with instructions from local skill files located in .agents/skills/.
Sanitization: There is no evidence of sanitization, escaping, or filtering applied to the diff output before it is passed to the next stage of processing.

polar-code-review