code
Pass
Audited by Gen Agent Trust Hub on Jun 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on local shell commands to manage the development lifecycle, including repository management via
git, code linting and testing viauv, and workspace operations through a customt3CLI utility. - [PROMPT_INJECTION]: An indirect prompt injection surface exists where the skill processes data from external tickets and URLs to guide code implementation.
- Ingestion points: Ticket descriptions and external URLs ingested via the
workspace ticketcommand and planning phase. - Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within ticket content.
- Capability inventory: Includes file writing/editing, shell command execution, and pull request management.
- Sanitization: No explicit sanitization of ticket-sourced content is described before it influences agent actions.
- [EXTERNAL_DOWNLOADS]: Performs remote repository synchronization (
git fetch) and provides informational links to external GitHub repositories such asgithub.com/obra/superpowers. - [DATA_EXFILTRATION]: Includes a mechanism to upload test evidence to a user-specified or tool-provided URL using the
lifecycle record-e2e-runcommand's--posted-urlparameter.
Audit Metadata