skills/yogsoth-ai/de-anthropocentric-research-engine/synthesize-meta-analytic-evidence/Gen Agent Trust Hub
synthesize-meta-analytic-evidence
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data in the form of
study_recordsand evidence synthesis tables. - Ingestion points: External data enters the agent context through the
study_recordsinput andextraction_tableoutputs defined in the contracts. - Boundary markers: The instructions lack explicit boundary markers or warnings to the agent to ignore potentially malicious content embedded within the study records or provenance maps.
- Capability inventory: The skill orchestrates a chain of sub-skills (e.g.,
design-data-extraction,audit-study-validity,assess-publication-bias) which aggregate and interpret data to make scientific decisions. While this file does not direct shell or network access, the interpolated data influences the agent's logic flow. - Sanitization: There is no evidence of specific sanitization or filtering logic for external content before it is processed into estimates or heterogeneity investigations.
Audit Metadata