audit-context-drift
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates exclusively on project documentation files (
CONTEXT.mdanddocs/adr/*.md) to ensure consistency. This is a standard and safe development task. - [DATA_EXFILTRATION]: There are no network operations or commands that could be used to exfiltrate data. The skill focuses on internal file reconciliation.
- [COMMAND_EXECUTION]: The skill does not execute shell commands or system-level processes. It uses high-level agent instructions for file reading and synthesis.
- [INDIRECT_PROMPT_INJECTION]: Although the skill processes external documentation files, it incorporates several security best practices, including using read-only sub-agents for analysis and a mandatory 'Verify before editing' step where the agent must confirm findings against the original ADR text before suggesting changes.
Audit Metadata