The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the analysis of local test files.
Ingestion points: The agent uses Read, Grep, and Glob to ingest the contents of .cs (C#) and .spec.ts (Playwright) files.
Boundary markers: There are no explicit delimiters or specific 'ignore instructions' markers defined for the content read from files.
Capability inventory: The skill is granted Write, Edit, and Bash permissions, which could be potentially misused if the agent follows malicious instructions embedded within code comments or strings in the files being audited.
Sanitization: No input sanitization or validation of the test file content is performed prior to processing.
Counter-measures: The skill includes strong behavioral instructions such as 'Be skeptical. Critical thinking. Everything needs traced proof' and 'Never accept code at face value', which provide a layer of protection against accidental obedience to injected instructions.

review-tests