agentic-evallisted

# Agentic Evaluation Patterns Patterns for self-improvement through iterative evaluation and refinement. ## Overview Evaluation patterns enable agents to assess and improve their own outputs, moving beyond single-shot generation to iterative refinement loops. ``` Generate → Evaluate → Critique → Refine → Output ↑ │ └──────────────────────────────┘ ``` ## When to Use - **Quality-critical generation**: Code, reports, analysis requiring high accuracy - **Tasks with clear evaluation criteria**: Defined success metrics exist - **Content requiring specific standards**: Style guides, compliance, formatting - **Iterative improvement workflows**: Drafts, plans, prompts, or code need critique and refinement before delivery ## Do Not Use This Skill For - Formal compliance claims, certification, or audit sign-off without approved organisational evidence - Production monitoring design, runtime governance, access control, or policy enforcement - Broad statistical evaluation methodology, benchmark design, or model selection studies - Security, privacy, or safety review where evaluation loops are not the central concern --- ## Expected Outputs When applying this skill, produce the assets needed to run and inspect an evaluation loop: - evaluation criteria or rubric - generator, evaluator, and optimiser responsibilities - iteration limit and convergence policy - structured evaluator output format - failure handling for malformed evaluations or n