agent-charter-operations
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides an operational framework for maintaining agent behavior standards through human-in-the-loop auditing. It does not contain any executable scripts, shell commands, or network requests.
- [NO_CODE]: The analyzed files consist entirely of Markdown documentation and structural templates. There are no programming language source files or binary executables.
- [INDIRECT_PROMPT_INJECTION]: The skill describes a process where agents ingest external data (pull request reviews and interrupt history) to propose behavioral rules. It correctly identifies this as a potential risk and mandates human review as a security gate:
- Ingestion points: Agent reads PR comments and interrupt logs (SKILL.md Section 1).
- Boundary markers: The framework specifies that agents only propose rules while humans approve them.
- Capability inventory: No execution capabilities are present in the skill; it only describes the methodology.
- Sanitization: All proposed changes must pass a manual weekly audit by a human curator before promotion.
Audit Metadata