orca-custom-framework
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFECREDENTIALS_UNSAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The instructions direct the agent to manually extract the Authorization token from the .mcp.json configuration file to authenticate API requests. This file typically contains sensitive credentials for the agent's environment.
- [COMMAND_EXECUTION]: The skill uses curl commands to perform create, update, and delete operations on the Orca Security API. This bypasses typical tool-based abstractions for state-changing operations and involves shell execution for network tasks.
- [INDIRECT_PROMPT_INJECTION]: The skill builds logic based on external data sources which could be manipulated to influence the agent's behavior.
- Ingestion points: Data is retrieved from the Orca environment using tools like discovery_search, get_alert, and get_compliance_framework_control_tests as described in SKILL.md.
- Boundary markers: The instructions do not specify any delimiters or safety prompts to prevent the agent from obeying instructions potentially embedded within alert descriptions or rule metadata.
- Capability inventory: The agent has the capability to write to the Orca API via curl to create frameworks and custom discovery alerts.
- Sanitization: There is no mention of filtering or validating the retrieved data before it is used to structure the custom frameworks.
Audit Metadata