update-hard-context
Pass
Audited by Gen Agent Trust Hub on Mar 31, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill manages personal context by reading and writing to a specific local file (docs/dashboard/advisor/hard.md). It does not perform any network operations, exfiltrate data, or execute untrusted code.
- [SAFE]: The update process includes a mandatory human-in-the-loop confirmation step where the agent must present a summary of changes to the user before writing to the file system or performing a git commit.
- [SAFE]: The ingestion of user-controlled file data for context updates follows best practices for task-specific context management.
- Ingestion points: Reads current status from the docs/dashboard/advisor/hard.md file.
- Boundary markers: Uses structured markdown headers to organize and delineate context sections.
- Capability inventory: Performs file read, file write, and git commit operations within the local repository.
- Sanitization: Relies on the user confirmation step to validate all generated updates before they are finalized.
Audit Metadata