The Agent Skills Directory

[COMMAND_EXECUTION]: Executes shell commands including git status, git diff, and npx playwright test to identify code changes and verify test execution. These are standard operations for a development and testing tool.
[PROMPT_INJECTION]: Contains internal instructions directing the agent to maintain a "fresh-eyes posture" and explicitly forbids skipping user confirmation steps (AskUserQuestion), even if prompted to work autonomously. These function as workflow constraints rather than malicious bypasses.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external files.
Ingestion points: Reads git diff output, local rule files (e.g., in .claude/rules/), test files, and frontend source code components.
Boundary markers: The skill does not define explicit delimiters or "ignore instructions" markers when reading these files.
Capability inventory: Possesses capabilities to execute shell commands (git) and run local test code (npx playwright).
Sanitization: No explicit sanitization or filtering is performed on the content of the audited files, creating a surface where malicious comments or strings in the source code could potentially influence agent behavior.

pw-test-audit