results-report
Pass
Audited by Gen Agent Trust Hub on May 13, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions and associated reference files contain no malicious patterns, remote code downloads, or exfiltration vectors. Its operations are limited to reading local research data and writing markdown reports to a dedicated results directory.\n- [PROMPT_INJECTION]: The skill includes an indirect prompt injection surface as it is designed to ingest and summarize the contents of external analysis files such as analysis-report.md and stats-appendix.md. This behavior is typical for report-generation skills and represents a vulnerability surface rather than an active exploit.\n
- Ingestion points: The skill reads analysis-report.md, stats-appendix.md, and figure-catalog.md as part of its standard workflow defined in SKILL.md.\n
- Boundary markers: No specific delimiters or warnings to ignore embedded instructions are used when reading input files.\n
- Capability inventory: The skill has the ability to write files to the Results/Reports/ directory and update project memory files.\n
- Sanitization: No input validation or sanitization is performed on the ingested content.
Audit Metadata