The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code and documentation from the local repository (git diffs, plans, PRDs, and source files). Malicious instructions embedded in these files could potentially influence the subagents performing the quality review.\n
Ingestion points: The skill reads git diff outputs, docs/plans/*.md, docs/prds/*.md, README.md, CLAUDE.md, and full content of staged source/test files (SKILL.md).\n
Boundary markers: There are no explicit instructions or delimiters defined to prevent subagents from interpreting embedded text in the reviewed files as commands or overrides.\n
Capability inventory: The skill spawns subagents using the Agent tool to perform complex text analysis and reporting across seven specific quality categories.\n
Sanitization: The ingested content is passed directly into subagent briefs without escaping or validation for malicious prompts.\n- [COMMAND_EXECUTION]: The skill executes several git commands to collect necessary inputs for the analysis, such as git diff --staged, git show, and git ls-files. These are standard operations for the skill's purpose but represent interaction with the local system environment.

implementation-review