The Agent Skills Directory

[SAFE]: Analysis of the skill instructions and structure confirms it is a legitimate tool for quality assurance and red-teaming. No malicious patterns, prompt injections, or unauthorized data access attempts were detected.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user documents and fetches data via web_fetch during Stage 1, creating an indirect prompt injection surface. This is evaluated as safe for this skill's use-case as its capabilities are limited to analytical feedback and standard search operations.
Ingestion points: Data enters via user-provided planning documents and external content retrieved through the web_fetch tool in Stage 1.
Boundary markers: No explicit instructions are provided to the agent to disregard instructions contained within the fetched content.
Capability inventory: The skill is limited to analysis and reporting using web_search and web_fetch. It contains no capabilities for file system modification, shell command execution, or code evaluation.
Sanitization: No specific sanitization, filtering, or validation of input content is specified.

adversarial-review