rolecard-workshop-ops
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements a 'read-only by default' posture and requires explicit user authorization before performing any state-mutating operations or deployments. It includes rigorous instructions for the redaction of sensitive data, such as OAuth secrets, private keys, and authorization headers, from all logs and communication channels. Guidelines for CORS policy and reverse proxy configuration further reinforce a security-first approach to infrastructure management.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external rolecard packages and metadata, which creates a potential surface for indirect prompt injection.\n
- Ingestion points: The agent ingests data from package-detail responses and imported packages during diagnosis and deployment sequences in SKILL.md and references/architecture-and-contracts.md.\n
- Boundary markers: The instructions mandate that all imported packages must pass schema, compatibility, and policy checks within a separate, previewable transaction before application.\n
- Capability inventory: The skill possesses capabilities for shell command execution via DEPLOY_COMMAND and VERIFY_COMMAND, as well as remote access via SSH_TARGET.\n
- Sanitization: Mandatory validation of schema, size, and content policy is required to ensure data integrity and prevent malicious payloads from influencing the deployment environment.
Audit Metadata