ao-workflow-runner
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill's setup instructions involve cloning repositories from GitHub (
github.com/jnMetaCode/agency-agents-zh.gitandgithub.com/jnMetaCode/agency-orchestrator.git) and installing a global NPM package (agency-orchestrator). These resources are provided by the skill author and are integral to the workflow runner's functionality. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data including YAML workflow files, Markdown role definitions, and user-supplied variables to drive the agent's behavior.
- Ingestion points: Reads
workflow.yaml,agency-agents-zh/{role}.md, and user inputs via theCollect Inputsstep. - Boundary markers: None explicitly defined to separate instructions from data during variable interpolation.
- Capability inventory: Performs local file reads to load roles and writes task outputs to the
ao-output/directory. - Sanitization: None specified; uses direct string replacement for
{{variables}}within task templates, which could allow instructions within data to influence the agent's persona or logic flow.
Audit Metadata