start-workflow

Pass

Audited by Gen Agent Trust Hub on Nov 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONPRIVILEGE_ESCALATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local Node.js scripts such as 'read-workflow-entry.mjs' and 'workflow-baseline.cjs' using arguments sourced from external manifests or user input, which could lead to command injection if inputs are not properly sanitized.
  • [INDIRECT_PROMPT_INJECTION]: The skill represents an indirect prompt injection surface by ingesting data from project-controlled files like 'workflows.json' and 'goal.md' to determine its execution sequence.
  • Ingestion points: 'workflows.json', 'goal.md', and user-provided workflow identifiers.
  • Boundary markers: No explicit delimiters are mentioned for data interpolation from the manifest into agent prompts.
  • Capability inventory: Execution of local scripts, spawning of sub-agents, and reading of local files.
  • Sanitization: The skill does not specify sanitization or validation logic for paths or identifiers retrieved from external manifests.
  • [DYNAMIC_EXECUTION]: It dynamically resolves and loads other skill files ('SKILL.md') at runtime based on instructions found in the workflow manifest.
  • [PRIVILEGE_ESCALATION]: Activation of a workflow automatically authorizes the 'spawn_agent' tool for sub-agents, which may bypass the standard per-request user consent for creating new agent instances.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 4, 2026, 10:12 AM