Rubber Duck Verifier
Break an execution loop by explaining the failure in text before touching code again. Use when stuck in a loop, guessing, or repeatedly failing tests — when the same file has been edited three or more times without the error changing, when fixes are being tried in sequence rather than chosen, when a test has failed the same way twice, or when a build or pipeline step keeps failing and each attemp…
Context Auditor
Clean the always-loaded context before a long build, so the agent is not reasoning from stale facts or rules that no longer apply. Use when starting a long or iterative task, when an agent keeps making the same wrong assumption, when it cites a file, count, path or version that does not match reality, when it refuses something reasonable for no visible reason, or when onboarding a project whose i…
Adversarial Reviewer
Review code as an adversary from a separate, clean context, so the agent that wrote it does not grade its own homework. Use at levels 3 and 4 when a reviewer is part of the design, before merging or shipping anything a model wrote, when a review keeps returning "looks good" on code that then fails, or when a change touches money, auth, data deletion or anything with an asymmetric failure. Also us…
Orchestration Design
Use when planning or fixing multi-step work, software or process: decide how much orchestration it needs, design it, build it on any runtime or none, and verify the build matches the design. SOFTWARE: multi-agent system, agent workflow, pipeline, batch job, ETL, RAG ingestion, CI/CD, LangGraph, StateGraph, fan-out/fan-in, supervisor/worker, reviewer/verifier agents, splitting one script into seve…