creating-experiments

Pass

Audited by Gen Agent Trust Hub on Sep 25, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it ingests untrusted data from external sources and interpolates it into critical tool calls.
  • Ingestion points: The skill reads project facts via the experiment-setup-context tool (detailed in SKILL.md Step 0 and references/setup-decisions.md) and accepts user-provided hypotheses and requirements.
  • Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when processing project data or user input for use in experiment descriptions or filters.
  • Capability inventory: The skill uses the ingested data to drive several high-impact tool calls, including experiment-create and experiment-update, which modify project state.
  • Sanitization: There is no evidence of sanitization or strict schema validation for the description (hypothesis) or property filter values before they are passed to the creation tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 25, 2026, 01:22 AM
Security Audit — agent-trust-hub — creating-experiments