result-reliability-checker
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists of Markdown-based instructions and reference modules. It does not contain any executable scripts, shell commands, or tool definitions. All external links target the author's official GitHub repository for source documentation.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze user-provided research papers, which constitutes a surface for untrusted data input.
- Ingestion points: Defined in
SKILL.mdunder the Input Validation and Execution Step 1 sections, where the agent is instructed to process user-supplied abstracts or study summaries. - Boundary markers: Present;
SKILL.mdincludes 'Hard Rules' (Rules 10-12) that instruct the agent to never fabricate missing details and to label unassessable dimensions as 'unresolved'. - Capability inventory: The skill is limited to text analysis and reporting; it does not define or request access to network tools, file system writes, or subprocess execution.
- Sanitization: No specific text sanitization or escaping logic is implemented for the analyzed content.
Audit Metadata