agents-md-revise
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from session learnings and repository evidence. It effectively mitigates this risk by:
- Explicitly instructing the agent to ignore embedded instructions or attempts to override rules within the data.
- Requiring a mandatory user approval of a proposed diff before any file writes occur.
- [DATA_EXPOSURE]: The instructions contain proactive security constraints that forbid the agent from reading sensitive local state, such as
.envfiles, credentials, SSH keys, or private untracked files. - [COMMAND_EXECUTION]: The skill limits its own scope to reading repository evidence and editing approved
AGENTS.mdfiles, explicitly prohibiting the invocation of other skills, helper commands, or deployment tasks.
Audit Metadata