workflow-runner

Pass

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to suggest downloading resources from external sources if not found locally. Specifically, it references a GitHub repository (github.com/jnMetaCode/agency-agents-zh.git) and an NPM package (agency-agents-zh).
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external YAML workflow files and Markdown role definitions to determine agent behavior. This creates a surface for indirect prompt injection where malicious content in those files could influence the agent's actions.
  • Ingestion points: YAML workflow files and Markdown role definition files (e.g., {agents_dir}/{role}.md) as described in Steps 1 and 4a.
  • Boundary markers: The skill uses {{variable}} templates for task rendering but does not specify explicit safety boundary markers or instructions to ignore embedded commands within the role files.
  • Capability inventory: The skill utilizes file reading (Read tool), file writing (Write tool for results in .ao-output/), directory checking (Bash tool), and sub-agent spawning (Agent tool).
  • Sanitization: No specific sanitization or validation of the content within the YAML or Markdown files is mentioned before processing or role-playing.
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to perform directory existence checks (test -d) to locate role definitions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 9, 2026, 06:31 PM
Security Audit — agent-trust-hub — workflow-runner