sdlc-design-story-tests
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes shell commands to interact with the repository, including
gitoperations (fetch, show, grep, branch management) and the execution of a project-local Python tool (.wiki-llm/tools/wiki_llm.py). These actions are scoped to the project wiki and follow a controlled drafting process. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the project wiki to generate test cases, presenting a potential surface for indirect prompt injection.
- Ingestion points: Markdown files in the
wiki/directory and contents of git branches. - Boundary markers: The skill implements a robust framework of 'gates' (Intent, Design, Change-set) requiring explicit human approval at multiple stages of the lifecycle.
- Capability inventory: File creation and modification within the workspace, git commit and branch operations, and execution of local repository scripts.
- Sanitization: Relies on manual human review during the defined approval gates rather than automated filtering.
Audit Metadata