The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it must ingest and analyze untrusted code content from the local workspace.\n
Ingestion points: SKILL.md directs the agent to read git diff output and the full contents of modified files.\n
Boundary markers: The skill does not implement delimiters or specific instructions to ignore embedded commands within the code being reviewed.\n
Capability inventory: The agent can execute git commands and is instructed to run project-specific quality checks like linters and tests.\n
Sanitization: No sanitization or validation of the ingested code content is performed.\n- [COMMAND_EXECUTION]: The skill executes shell commands as part of its core workflow.\n
It uses git diff to collect context.\n
It directs the agent to execute arbitrary tools defined in the local project environment (e.g., lint, type, tests), which could be a risk if the project configuration is compromised.

review