cs-agent-setup
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses npx, npm, git, curl, and other shell tools to inspect and configure the local environment and the @codespring-app/use-agent SDK. These commands are used to list agents, verify authentication status, and fetch skill templates.
- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external data through WebSearch and WebFetch tools and reads local repository files. This creates a surface where instructions embedded in external content or code could potentially influence the agent's behavior.
- Ingestion points: WebSearch results, WebFetch content, and repository files accessed via rg and sed.
- Boundary markers: The instructions define a "Two-layer rule" for user communication and require a setup receipt, but do not specify technical delimiters to isolate tool outputs from agent instructions.
- Capability inventory: Shell execution via Bash (npx, npm, git, curl) and file system access.
- Sanitization: The skill relies on instructional constraints rather than programmatic sanitization of external data before processing.
- [SAFE]: The skill implements strong secret management policies. It explicitly instructs the agent not to ask for provider keys or API secrets in the chat and guides the user to use secure methods like interactive CLI login and dashboard forms.
Audit Metadata