squad-run
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
curlto communicate with an external API ($BASE_URL) for fetching task requirements and updating pipeline status. This is a fundamental component of the skill's orchestration functionality. - [COMMAND_EXECUTION]: The skill executes system commands including
gitfor version control operations (commit, add, log) andpython3for rendering complex agent prompts from templates and processing JSON data. - [PROMPT_INJECTION]: The skill processes task titles and descriptions from an external API, creating a potential surface for indirect prompt injection.
- Ingestion points: Data enters the context via API GET requests for task metadata.
- Boundary markers: External data is interpolated into sub-agent prompts using template placeholders like
<title>and<description>. - Capability inventory: The system can perform repository writes via Git and update external task states via API calls.
- Sanitization: The skill relies on a multi-stage review process (Critic and Inspector agents) and a dedicated prompt-rendering script to validate and structure sub-agent inputs.
Audit Metadata