arch-plan
Pass
Audited by Gen Agent Trust Hub on Mar 31, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill follows secure development practices by prioritizing local implementation and requiring explicit confirmation for new architectural plans. No evidence of obfuscation, hardcoded credentials, or unauthorized persistence was found.- [PROMPT_INJECTION]: The skill processes external information, which creates an inherent surface for indirect prompt injection. \n
- Ingestion points: The skill ingests untrusted data from existing documentation in the
docs/directory and from web content retrieved during the "external research" phase. \n - Boundary markers: Document integrity is maintained using stable block markers (e.g.,
arch_skill:block:phase_plan) to delimit different sections of the architecture plan. \n - Capability inventory: The agent has the capability to write to the file system (documentation and worklogs) and perform local code modifications during the execution phase. \n
- Sanitization: The skill relies on the agent's synthesis of research data; while it does not define explicit sanitization routines, it mandates that claims must be anchored in file paths and symbols, which acts as a verification step.
Audit Metadata