skills/soujava/agent-skills/sldd/Gen Agent Trust Hub

sldd

Pass

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is designed to execute shell commands to run test suites during the Red (failing tests) and Green (passing tests) development phases, as detailed in steps/feature/04-tests-red.md and steps/feature/05-implementation-green.md. This behavior is central to the skill's purpose as a software development tool.
  • [INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection by processing external data. 1. Ingestion points: User-provided ideas in steps/feature/88-exploration.md and project repository files in steps/feature/99-codebase-context.md. 2. Boundary markers: The skill employs an extensive gate-based architecture and separates instructions into specific files per step to enforce logical boundaries. 3. Capability inventory: File system read/write access and shell command execution for testing (Steps 04 and 05). 4. Sanitization: The workflow relies on mandatory human-in-the-loop approval protocols before any artifacts are persisted or implementation logic is executed.
  • [DYNAMIC_EXECUTION]: The skill facilitates the generation of production code and executable tests, which are then run by the agent (Steps 04 and 05). This runtime script creation and execution cycle is the primary intended function of the Spec Loops Driven Development methodology.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 6, 2026, 12:38 AM