evidence-accumulation-selector

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill consists entirely of instructional markdown and mathematical formulas related to cognitive science. There are no executable scripts or system commands present in the files.
  • [EXTERNAL_DOWNLOADS]: The skill mentions various academic software tools (HDDM, fast-dm, PyDDM, etc.) and R packages (rtdists, DMC) used in the field. These are standard research tools from academic sources and well-known repositories, which are documented neutrally for informational purposes.
  • [INDIRECT_PROMPT_INJECTION]: While the skill is designed to process user-provided experimental data (accuracy and reaction times), it lacks any capability to execute code or perform network requests. Consequently, there is no viable attack surface for data-driven injection to escalate privileges or exfiltrate information.
  • [METADATA_POISONING]: The metadata fields (author, version, description) are consistent with the instructional content and match the author's established repository context without deceptive patterns.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 11:51 AM
Security Audit — agent-trust-hub — evidence-accumulation-selector