bitfsae-project-standards
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and audit existing repository content, including README files, source code, and design documents. This creates a surface for indirect prompt injection if those files contain malicious instructions. However, the skill includes explicit safety boundaries such as 'Do not change intended behavior during a documentation-only task' and requirements to verify all claims against the implementation, which mitigates the risk.\n
- Ingestion points:
SKILL.mdworkflow steps 1-3 involve reading the repository tree, source files, and build configurations.\n - Boundary markers: The skill requires distinguishing verified behavior from design intent and explicitly marking unverified hardware mappings.\n
- Capability inventory: The skill allows the agent to read and write repository documentation and configuration files.\n
- Sanitization: The skill emphasizes cross-referencing claims across multiple authoritative sources (code, generator configs, hardware files) to ensure accuracy.\n- [DATA_EXFILTRATION]: The skill provides explicit safety instructions regarding sensitive data, stating 'Keep secrets out of public repositories' and 'Do not expose real credentials or private keys' in the documentation system guidelines.\n- [REMOTE_CODE_EXECUTION]: While the skill mentions external repositories (e.g.,
BITFSAE/vehicle-interfaces), these are referenced as sources of truth for documentation and manual review, not for automated code execution or unverified dependency installation.
Audit Metadata