Routes users to the appropriate sdd-lite evaluation level based on testing goals.
npx skills add https://github.com/nico0695/ai-tools --skill sddl-eval-nico0695
Please help me install this Agent Skill. Skill: sddl-eval Source: https://github.com/nico0695/ai-tools/tree/main/sdd/sdd-lite-evals/skills/sddl-eval Command: npx skills add https://github.com/nico0695/ai-tools --skill sddl-eval-nico0695
sdd-lite-evals workspace already exists and directs users to sddl-eval-init if not. - Level Routing: Maps user intent to the right skill: structural confidence to level 1, trigger/routing/worker behavior to level 2, real proposal-to-QA behavior to level 3, and history comparison to the report skill. - Cost Transparency: Explains expected model calls, mutations, and cost class (none, moderate, or high) before handing off. - Use Case: A user asks how to validate that their sdd-lite installation triggers workers correctly; the navigator recommends sddl-eval-level-2 with a moderate cost estimate and lists what information to prepare. ## Quick Start Ask which sdd-lite evaluation level you should run to test your installation's routing behavior.High-intent search queries and answers about installing and using this skill.
Match your goal to a level: level 1 for structural or installation confidence, level 2 for trigger, wrapper, routing, worker, or resume behavior, and level 3 for real proposal-to-QA behavior. Use the report skill for history or baseline comparison.
First confirm a sdd-lite-evals workspace exists for the project; if not, run sddl-eval-init. Then pick the evaluation level matching your concern and review the expected model calls and mutations before starting.
Level 1 checks structural and installation confidence, level 2 exercises trigger, routing, worker, and resume behavior, and level 3 runs real proposal-to-QA behavior. Higher levels involve more model calls and higher cost.
Yes, most levels invoke model calls, and the navigator reports an expected cost class of none, moderate, or high before handoff. It never launches model-backed child sessions itself, so costs only occur in the selected level skill.
Skip the navigator when you already know which level to run or when no sdd-lite installation exists to evaluate. It is a routing layer only and does not perform the evaluation itself.
nico0695Bootstraps sdd-lite project context, skill catalog, and AI agent configuration files.
nico0695Reviews execution stages or completed changes and produces a qa-report.md with verdict and next action.
homeassistant-aiCompare MCP tool behavior across versions using diff-based story evaluation.
leeoveryGenerate a feature specification markdown file from inline context.