Browse 26 Deterministic Verification Loops
YAML state machines and Python/Rust loop controllers with retry budgets, bash execution, and exit-code gates. 26 free and paid blueprints with pre-filled examples.
Deterministic Verification Loops blueprints
Engineering Loop: SWE-Agent Autonomous Bash & Test Verification State Machine
Standard SWE-Agent state-machine loop configuration executing commands in an interactive bash subshell, verifying exit codes, capturing compiler/test errors, and updating state until tests pass.
Engineering Loop: SWE-Agent Pure Bash Tooling & Diagnostic Repair Loop
Minimalist SWE-Agent loop operating without custom editor binaries, relying solely on POSIX bash utilities, patch, sed, and git diff.
Engineering Loop: Mini SWE-Agent Lightweight Python Verification State Machine
Ultra-lightweight state machine for SWE-Bench evaluation featuring minimal token overhead, single-call tool dispatch, and deterministic step counter.
Engineering Loop: SWE-Agent Markdown Backticks Code Replacement Loop
SWE-agent loop variant using fenced markdown code blocks with line numbering for deterministic multi-line block replacements.
Engineering Loop: SWE-Agent Multimodal Visual Inspection & UI Verification Loop
Multimodal state machine capable of capturing webpage or GUI screenshots, rendering visual diffs, and verifying visual layouts against test specifications.
Engineering Loop: Mini SWE-Agent SWE-Bench Benchmark Evaluation State Machine
Production benchmark loop calibrated for SWE-bench Lite and Verified test suites with automated git worktree isolation and test isolation.
Engineering Loop: LangGraph Code Generation, Import Checking & Pytest Self-Correction Graph
Deterministic LangGraph state machine implementing: Generate Code -> Check Imports -> Run pytest/ruff -> Conditional Edge (Finish if exit code 0, else loop back with traceback).
Engineering Loop: LangGraph Reflexion Deterministic State Machine (Actor-Evaluator-Reviser)
Deterministic implementation of the Reflexion architecture in LangGraph: Actor produces output -> Evaluator scores against ground truth/rubric -> Reviser reflects on critique -> Conditional loop back until quality threshold is met.
Engineering Loop: AlphaCodium Synthetic Edge-Test Generation & Self-Validation Stage
Stage dedicated to generating high-coverage synthetic unit tests covering edge conditions, corner cases, empty inputs, and maximum bounds.
Engineering Loop: AlphaCodium Test Failure Analysis & Iterative Code Fix Loop
Iterative repair loop separating the diagnosis phase from the fix phase: first analyzes why the code failed on input X, then applies targeted surgery.
Engineering Loop: Loop-Engine Goal-Directed Deterministic Controller
Pure-Python goal-directed loop controller demonstrating failure isolation, retry budgets, monotonic state progress, and stop conditions without LLM hallucination drift.
Engineering Loop: Loop-Engine Directed Acyclic Graph (DAG) State Transition Engine
Graph-based state machine defining nodes, typed transitions, validation predicates, and fallback edges for autonomous agent workflows.
Engineering Loop: Browser-Use Agent Action-Observation Playwright Execution Loop
Core action-observation state machine for autonomous web browsing using Playwright: DOM extraction -> Visual Element Tree -> Action Execution -> Result Verification.
Engineering Loop: Stagehand Interactive Action Observation & Element Mutation Loop
Stagehand observe-and-act loop where available actions are dynamically observed, scored for relevance, and executed with automatic locator recovery.
Engineering Loop: Browser-Use LLM-as-a-Judge Extraction & Goal Completion Verifier
Independent judge loop verifying whether scraped data matches the user's intent and conforms to the requested Pydantic schema before terminating.
Engineering Loop: Zeroshot Rust Autonomous Goal Supervisor & Run Ledger
High-performance Rust autonomous supervisor implementing atomic checkpoint commits, delivery gates, and hard resource limit enforcements.
Engineering Loop: SWE-Agent Competitive Programming & Edge-Case Loop
State machine optimized for competitive programming benchmarks with strict timeout bounds, edge test generation, and stdout matching.
Engineering Loop: AlphaCodium Multi-Stage Test-Driven Code Generation Flow
Full AlphaCodium multi-stage architecture: Problem Reflection -> AI Test Generation -> Public Test Run -> Iterative Code Fix on Failing Tests -> Multi-Candidate Selection.
Engineering Loop: Stagehand Paginated Catalog Extraction & Fault-Tolerant Checkpoint Loop
Production cookbook loop extracting multi-page e-commerce catalogs with local JSON checkpointing, next-page button healing, and pagination termination logic.
Engineering Loop: OpenHands CodeAct Iterative Bash & Python Execution State Machine
Foundational CodeAct state machine where the LLM writes executable code snippets and bash commands directly to perform multi-step software engineering repairs.
Engineering Loop: Zeroshot Diagnostic Post-Mortem & Failure Isolation Engine
Rust failure isolation engine that analyzes crashed agent sessions, parses crash dumps and tail logs, and classifies failures into fatal vs recoverable.
Engineering Loop: OpenHands Cyclic Repetition & Loop Stagnation Breaker
Deterministic pattern analyzer that inspects the rolling history of actions and observations to detect infinite loops, repeating commands, and non-advancing edit cycles.
Engineering Loop: OpenHands Automated Post-Generation Verifier & Diff Review Loop
Specialized subagent loop that verifies whether another agent's proposed changes actually solve the problem without introducing regressions or code style violations.
Engineering Loop: OpenHands AgentController Action-Observation Lifecycle Machine
Central coordinator state machine of OpenHands driving turn-by-turn agent execution, state persistence, delegate subagent coordination, and error containment.