sdlc-design-feature-tests
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the project wiki to generate test designs. This ingestion surface could potentially be exploited by malicious content within requirements or design files.
- Ingestion points: Markdown documentation files in the
wiki/directory including Requirements, Architecture, and Endpoints. - Boundary markers: Relies on structured schema definitions in
.wiki-llm/schema/to define context, though explicit instruction delimiters are not provided. - Capability inventory: File system write operations for creating test cases and Git operations for committing changes to the wiki repository.
- Sanitization: Risk is mitigated by a mandatory human-in-the-loop protocol requiring explicit approval at the Intent, Design, and Change-set gates, alongside an automated self-review by a sub-agent.
- [COMMAND_EXECUTION]: The skill executes local shell commands to manage the project wiki and validate its state.
- Evidence: Usage of standard utilities like
git,grep, andls, as well as a local Python scriptpython3 .wiki-llm/tools/wiki_llm.pydefined as a tool for checking wiki integrity and managing draft worktrees.
Audit Metadata