audit-agent-lifecycle
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the shell to run pre-existing Node.js utility scripts, specifically
update-agent-map.mjsandpopulate-agent-descriptions.mjs, to manage agent metadata and export findings. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted agent definition data, creating an attack surface for indirect prompt injection.
- Ingestion points: Programmatically reads all agent configuration and prompt files within the
agents/directory. - Boundary markers: The skill does not employ explicit delimiters or system instructions to prevent the model from following commands potentially embedded in the audited agent files.
- Capability inventory: The skill is capable of executing shell commands via Node.js, writing files to the
.plugin-creator/audits/directory, and dispatching sub-agents for file refactoring. - Sanitization: No sanitization, escaping, or filtering is performed on the content parsed from the agent files before it is processed or used in further instructions.
- [DATA_EXFILTRATION]: The skill collects detailed architectural data, including agent tool sets, internal instructions, and configuration contradictions, and centralizes this information into a JSON map and multiple markdown reports within the local
.plugin-creator/audits/directory.
Audit Metadata