agentic-engineering

--- name: agentic-engineering description: Operate as an agentic engineer using eval-first execution, decomposition, and cost-aware model routing. metadata: origin: ECC --- # Agentic Engineering Use this skill for engineering workflows where AI agents perform most implementation work and humans enforce quality and risk controls. ## Operating Principles 1. Define completion criteria before execution. 2. Decompose work into agent-sized units. 3. Route model tiers by task complexity. 4. Measure with evals and regression checks. ## Eval-First Loop 1. Define capability eval and regression eval. 2. Run baseline and capture failure signatures. 3. Execute implementation. 4. Re-run evals and compare deltas. ## Task Decomposition Apply the 15-minute unit rule: - each unit should be independently verifiable - each unit should have a single dominant risk - each unit should expose a clear done condition ## Model Routing - Haiku: classification, boilerplate transforms, narrow edits - Sonnet: implementation and refactors - Opus: architecture, root-cause analysis, multi-file invariants ## Session Strategy - Continue session for closely-coupled units. - Start fresh session after major phase transitions. - Compact after milestone completion, not during active debugging. ## Review Focus for AI-Generated Code Prioritize: - invariants and edge cases - error boundaries - security and auth assumptions - hidden coupling and rollout risk Do not waste review cycles on style-only disagreements when automated format/lint already enforce style. ## Cost Discipline Track per task: - model - token estimate - retries - wall-clock time - success/failure Escalate model tier only when lower tier fails with a clear reasoning gap.

[{"step":"Define the problem and success criteria","action":"Clearly state the task, constraints (e.g., cost, time), and measurable outcomes. Use [PLACEHOLDERS] to standardize inputs.","tip":"Include specific metrics (e.g., 'reduce latency to <500ms') to guide the agent's evaluation."},{"step":"Decompose the task","action":"Break the problem into 3-5 sub-tasks with clear success criteria for each. Prioritize tasks that are prerequisites for others.","tip":"Use tools like Mermaid diagrams or bullet points to visualize dependencies between sub-tasks."},{"step":"Select models/tools for each sub-task","action":"Use cost-aware routing to pick the best model/tool for each step. For example, use smaller models for initial exploration and larger ones for complex validation.","tip":"Track model costs in a spreadsheet or tool like LangSmith to optimize spending over time."},{"step":"Execute and evaluate iteratively","action":"Run each sub-task, evaluate against criteria, and adjust if needed. Document failures and refinements in a log.","tip":"Use eval frameworks (e.g., pytest, custom scripts) to automate validation where possible."},{"step":"Iterate and optimize","action":"After completing all sub-tasks, review the full solution for further optimizations (e.g., combining steps, reducing model calls).","tip":"Use the agent to simulate edge cases or stress tests to uncover hidden bottlenecks."}]