rls-policies
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data which could potentially contain malicious instructions intended to manipulate the agent's behavior.
- Ingestion points: The agent reads application source code (via
parameters.dataApp.git), README files, configuration text, and sampled database column values as part of the discovery process. - Boundary markers: The skill includes a dedicated 'Untrusted input' section in
SKILL.mdthat explicitly instructs the agent to treat this content as data only and ignore any embedded instructions. - Capability inventory: The agent can execute
kbagentcommands to create, update, or delete security policies and modify data app configurations. - Sanitization: The skill requires the agent to present a summary of all intended changes and wait for an explicit user confirmation ('go') before performing any write operations.
- [COMMAND_EXECUTION]: The skill utilizes the
kbagentCLI and standard shell utilities to perform administrative tasks. - Evidence: Instructions in
references/kbagent.mddescribe usingkbagentfor policy management and configuration updates, as well as shell utilities likeuuidgenandmktempfor session and file management. These operations are gated by dry-run steps and user confirmation requirements.
Audit Metadata