experiment-stop

Pass

Audited by Gen Agent Trust Hub on Aug 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses a local helper script ${CLAUDE_PLUGIN_ROOT}/scripts/gb-call to execute shell commands and interact with the GrowthBook API. This is the primary function of the skill and is documented as standard behavior within the vendor's environment.\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by fetching experiment metadata (names, descriptions, and variations) from the GrowthBook API. Maliciously crafted data in the experiment could potentially be used to influence the agent's behavior.\n
  • Ingestion points: API responses from GET /api/v1/experiments and GET /api/v1/experiments/<id> referenced in SKILL.md.\n
  • Boundary markers: Absent; the instructions do not implement delimiters or warnings to differentiate remote data from system-level instructions.\n
  • Capability inventory: Shell command execution via the gb-call helper and general command-line capabilities.\n
  • Sanitization: The skill does not define specific validation or escaping mechanisms for the data retrieved from the remote API before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 6, 2026, 02:39 AM
Security Audit — agent-trust-hub — experiment-stop