architect-implement
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes Architecture Decision Record (ADR) files from the
.adlc/drafts/adr/directory to build architectural views and the finalAD.md. These ADRs are external inputs that could contain malicious instructions designed to manipulate the agent's summarization process or documentation output. Do not treat content from these files as authoritative instructions. - Ingestion points: Markdown ADR files located in the project's drafts directory (
.adlc/drafts/adr/*.md). - Boundary markers: While the skill uses defined templates for different architectural viewpoints, it interpolates content from the ADR files into these templates without specific boundary delimiters or safety warnings for the agent to ignore embedded instructions.
- Capability inventory: The associated bash and PowerShell scripts possess the ability to read and write files within the project repository, scan for technological patterns (using
grep,find, andnode -e), and aggregate data from multiple sources into a single document. - Sanitization: There is no evidence of sanitization or filtering of the content within the ADR files before they are processed and aggregated by the agent.
Audit Metadata