autopilot
Pass
Audited by Gen Agent Trust Hub on Mar 16, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes shell commands to manage a local development environment during the QA phase. It uses
lsofto check for services on port 3000,bun dev:nextto initiate a development server, andkillto terminate processes. These are standard operations for an autonomous development agent.\n- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it processes untrusted data from GitHub issues via thegh issue viewcommand. This content is used to inform code generation and project specifications.\n - Ingestion points: Untrusted issue content is retrieved using
gh issue viewinSKILL.md.\n - Boundary markers: The skill does not employ explicit delimiters or instructions to isolate issue descriptions from the agent's internal logic.\n
- Capability inventory: The agent can execute shell commands, perform local network requests, and commit code to a repository via associated helper skills.\n
- Sanitization: There is no evidence of input validation or sanitization applied to the retrieved GitHub data before it is processed by the AI models.
Audit Metadata