devcodex-labs/devcodex

AI Agent System Architecture covers agent routing, tool permissions, context, memory, state, and observability for tasks that need explainable and auditable agent behavior. Use it when designing agent workflows, recovery, handoff, or model-assisted governance.

AI Evaluation Engineering designs evidence for model and prompt quality using golden sets, rubrics, judge calibration, repeat sampling, variance reporting, and regression decisions. Use it when comparing models, prompts, or upgrades and you need reproducible results with contamination checks.

API Contract Architecture defines and freezes public API contracts before implementation. Use it for HTTP, SDK, CLI, schema, error, pagination, idempotency, and compatibility changes that affect consumers.

Accessibility I18n handles accessibility and localization review for user-facing UI, CLI output, and docs. Use it when tasks involve keyboard flow, focus, screen readers, ARIA, RTL, translations, locale formats, or multilingual content.

Analyze Default defines a read-only workflow for multi-round code analysis, evidence gathering, and convergence checks. Use it when you need a factual conclusion, root-cause assessment, or review of whether something is correct without changing files.

Analyze Research handles technical selection, feasibility checks, dependency comparisons, and root-cause investigation in a read-only workflow. Use it when you need evidence-based recommendations without changing source files.

Api Verification generates archive-grade .http examples and .cjs automation scripts for API changes. Use it when validating endpoint behavior, contract fields, or regression checks in dev and fix workflows.

Audit Common enforces the shared gates used by all audit workflows, including G0 scale routing, profile freshness checks, and G1~G5 review dimensions. It is used before subtype-specific checks like TD, RQ, PE, RA, DA, and RL.

Audit Dimensions reviews AI instruction files and specs against 25 dimensions such as structure, content, routing, and markdown correctness. Use it when checking skills, instructions, agents, or rules documents for execution-ready issues.

Audit Document reviews general project documents such as READMEs, architecture docs, and developer guides. Use it to check structure, accuracy, links, terminology, audience fit, and consistency.

Audit Execution Guide defines how to run audit work in prioritized batches, route targeted review subsets, and apply linked checks for new subtypes, dimensions, and numeric references. It is used when coordinating multi-stage code or document audits.

Audit Project reviews project engineering quality across structure, dependencies, testing, security, and resource lifecycle risks. Use it when assessing codebases, directories, and build health for maintainability and leaks.

Audit Readme reviews README files for user paths, quick start steps, example realism, configuration and troubleshooting visibility, and consistency with other entry points. Use it when the README is the main user-facing entry for a project or tool.

Audit Report reviews report documents for completeness, factual accuracy, traceability, and actionable findings. Use it when auditing analysis, review, development, or fix reports, especially external review recaps.

Audit Requirements reviews requirement documents for completeness, clarity, consistency, traceability, constraints, and change history. Use it when checking whether a PRD or requirement brief is ready for technical design and validation planning.

Audit Session persists audit rounds, findings, and convergence state in a session JSON file so an audit can resume after interruption. Use it for multi-round audit workflows that need exact recovery across sessions.

Audit Tech Design审查技术方案、架构设计、技术选型和实施文档的质量与风险。用于检查边界、兼容性、安全性、测试和文档一致性。

Backend Domain Architecture guides backend design from domain language, business rules, and consistency boundaries before implementation details. Use it for APIs, services, permissions, transactions, idempotency, and compatibility changes.

Brand Visual Quality verifies brand marks, icons, theme variants, and micro-size outputs against a traceable master. It is used when you need geometry parity, mono-master checks, and visual evidence for acceptance.

Business Model Review checks commercial assumptions around pricing, packaging, revenue, costs, and operating burden. Use it when a change affects monetization, paid paths, channels, or long-term sustainability.

CP Gate enforces CP1 requirement confirmation, CP2 solution confirmation, and conditional CP3 implementation-plan confirmation. It is used to control workflow approval order before an agent starts work.

Compliance executes entry checks and FC/SC/RC/T policy validation for Claude Code workflows. Use it when you need mode-aware compliance handling, with full validation in dev and chat exemptions for the compliance block.

Consumer Validation Engineering verifies that source changes are reproducibly tested by real consumers across repositories. It is used for SDKs, CLIs, frameworks, and public packages when identity, packed artifacts, fresh installs, CI linkage, and denominator-based release claims must be proven.

Release Audit reviews release readiness, changelog quality, compatibility and migration risk, package metadata, docs sync, rollback plans, and registry or tag risks before publication. It is used for pre-release review and risk checks, not for performing the release.

User Manual Audit reviews user-facing documentation, README files, and doc sites for clarity, structure, navigation, and first-time success. Use it when checking whether end-user docs are understandable and usable.