AoA Tech Library: Reusable Engineering Practice Specifications for Coding Agents and Developers

aoa-techniques is an open-source library of engineering practice specifications that provides reusable, verifiable, and orchestratable atomic technical units for coding agents and human developers, covering core areas such as verification patterns, security protocols, document structures, and evaluation loops. This project is not just a collection of code snippets but a carefully designed public library of engineering practice specifications with dual attributes: serving as an independent public library for developers to integrate individual technical units, and acting as an internal component of AoA with stable IDs and topological structures.

aoa-techniques (Agents of Abyss Tech Library) is a unique open-source project whose core positioning is to provide reusable, verifiable, and orchestratable atomic technical units for coding agents and human developers. Core philosophy: A technical unit should be an atomic, executable, and reproducible engineering practice that can be classified, templated, verified, and delivered to small agents in the correct context. Its dual attributes: 1. Independent public library: Developers can extract individual technical units to integrate into agent systems without deploying the entire OS Abyss ecosystem; 2. Internal component of AoA: Within OS Abyss, it maintains stable IDs, topological structures, sources, and review mechanisms, generating interfaces consumable by sibling repositories.

aoa-techniques has strict definitions for technical units, distinguishing them from random code snippets or project-specific hacks. An excellent technical unit must include: clear intent and usage boundaries, atomic executable actions, topological adaptation across domain/kind/future axes, explicit input/output and risks, verification methods, portability adaptation instructions, and sufficient structure to support promotion, review, and reuse. Content not suitable for inclusion: random code snippets, private project hacks without adaptation instructions, configurations containing sensitive information, raw logs, dumps applicable only to specific projects, undocumented scripts, and objects more suitable as skills/evaluations/routing logic/role contracts/scenarios.

aoa-techniques uses a carefully designed directory structure to ensure discoverability and maintainability. Core directories include: techniques/ (released technical unit bundles), templates/ (creation scaffolding), generated/ (derived directories, capsules, etc.), legacy/ (historical versions and migration records), mechanics/ (cross-mechanism movement interfaces), docs/ (directions, review guidelines, etc.), scripts/ and tests/ (auxiliary tools). Technical unit statuses: promoted (promoted), canonical (canonical), deprecated (deprecated). Early incubation occurs elsewhere, with the root directory pointing to selected public specifications.

aoa-techniques covers key areas of coding agent engineering: 1. Agent workflows (plan-execute-verify loops, multi-agent collaboration protocols, etc.); 2. Verification patterns (code correctness verification, output format validation, etc.); 3. Document structures (README templates, API documentation specifications, etc.); 4. Evaluation and monitoring loops (performance metrics, quality assessment frameworks, etc.);5. Security and hygiene protocols (input validation, sensitive information handling, etc.);6. Infrastructure operation techniques (configuration management, service discovery, etc.);7. Cross-repository promotion and reuse patterns (dependency management, interface adaptation, etc.).

aoa-techniques adopts a unique contribution model: separation of technical ownership and execution responsibility—aoa-techniques owns the practical meaning (defining what a technique should do and how to verify and classify it), while sibling repositories own execution, proof, etc. The process from candidate to release for a technique: pick→inspect→expand→object use. Migration mechanism: When an atomic reusable contract can be clearly extracted from a sibling repository and meets public safety, clear boundaries, and portability requirements, it can be included in aoa-techniques.

aoa-techniques explores how to systematize and standardize human engineering wisdom so that it can be understood and executed by agents. For agent developers: It provides reusable best practices, verified patterns, and standardized interfaces; For human developers: Knowledge precipitation, collaboration foundation, and quality assurance; For the industry: It provides engineering solutions for making the behavior of AI coding agents predictable, verifiable, and auditable—building reusable technical infrastructure, decomposing intelligence into verifiable atomic units, and establishing clear boundaries for human-machine collaboration.

The project provides multiple quick paths: by requirement type (repository ownership, tech map, etc.), by technical attribute (atomic contracts, function readers, etc.), by review and promotion status (status interface, promotion readiness, etc.), end-to-end examples (WALKTHROUGH.md), boundary execution and proof neighbors (aoa-skills, aoa-evals, etc.). For details, refer to documents such as CHARTER.md, TECHNIQUE_INDEX.md, and docs/VIA_NEGATIVA_CHECKLIST.md.