product-stress-test

Pass

Audited by Gen Agent Trust Hub on Mar 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local Go scripts and Git commands for state assessment, data retrieval, and version control (e.g., ../../scripts/cmd/assess, ../../scripts/cmd/compound). It utilizes shell variables like $SLUG in command lines, which relies on platform-level sanitization to mitigate command injection risks.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes untrusted content from PRDs and value propositions to influence the behavior of sub-agents.\n
  • Ingestion points: Reads files from .agents/products/prds/ and .agents/products/value-propositions/.\n
  • Boundary markers: The instructions do not define explicit boundary markers or delimiters to isolate document content from agent instructions.\n
  • Capability inventory: The agent has capabilities to execute shell commands (go run, git) and modify files on the local filesystem.\n
  • Sanitization: There is no evidence of sanitization or validation of the ingested file content before it is passed to evaluator sub-agents.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 27, 2026, 06:43 PM