The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8). It reads the content of an external target file (a skill document) and passes either the content or the file path to multiple specialized reviewer sub-agents. A malicious skill document could contain embedded instructions designed to influence the behavior of these reviewers or the subsequent automated editing phase.
Ingestion points: lib/001_INITIALIZE.md (validates and reads target file) and lib/002_FAN_OUT.md (passes path to sub-agents).
Boundary markers: None detected. The reviewer prompts in lib/prompts/ do not use delimiters or instructions to ignore potential injections within the target file.
Capability inventory: The skill utilizes the Edit tool to modify files (lib/007_ADDRESS.md) and the Bash tool to perform version control operations including git add and git commit (lib/010_STAGE.md, lib/011_COMMIT.md).
Sanitization: No explicit sanitization or filtering of the target file's content is performed before it is processed by the agent or its sub-tasks.
[COMMAND_EXECUTION]: The skill makes extensive use of system commands via the Bash tool to interact with git. While these operations are core to the skill's purpose (automated code review and committing), they represent a capability that could be targeted via the prompt injection surface mentioned above. The skill does implement a secure heredoc pattern for commit messages in lib/011_COMMIT.md to mitigate shell injection risks.

review-skill