Agent skill

AI Agent System Architecture

by devcodex-labs in devcodex-labs/devcodex

AI Agent 系统架构专家 Owner — 当任务涉及 Agent 路由、工具调用、上下文管理、记忆、状态机、权限、人机协作、可观测性、回放验证或模型辅助治理时使用;要求把 Agent 行为设计成可解释、可恢复、可审计。

AGPL-3.0Auto-check passed

Install AI Agent System Architecture

skills CLI
$ npx skills add devcodex-labs/devcodex --skill ai-agent-system-architecture -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install devcodex-labs/devcodex ai-agent-system-architecture --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/devcodex-labs/devcodex.git skills-src && mkdir -p .claude/skills && cp -r skills-src/content/skills/ai-agent-system-architecture .claude/skills/ai-agent-system-architecture && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
ai-agent-system-architecture
GitHub stars
439
Token cost
~2.4k tokens
SKILL.md length
572 words
Files
2
Skills in repo
70
Repo updated
First seen
Licence
AGPL-3.0

At a glance

AI Agent 系统架构专家 Owner — 当任务涉及 Agent 路由、工具调用、上下文管理、记忆、状态机、权限、人机协作、可观测性、回放验证或模型辅助治理时使用;要求把 Agent 行为设计成可解释、可恢复、可审计。

  • Works in 5 steps: 建立 Agent… → 明确工具权限边界:宿主拥有文件、删除和命令权限,DevCodex… → 定义上下文、记忆、handoff、summary 和恢复优先级。 → …
  • SKILL.md covers 定位, 触发条件, 核心门禁 and 执行步骤, plus 3 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

AI Agent System Architecture is an agent skill from devcodex-labs/devcodex. AI Agent 系统架构专家 Owner — 当任务涉及 Agent 路由、工具调用、上下文管理、记忆、状态机、权限、人机协作、可观测性、回放验证或模型辅助治理时使用;要求把 Agent 行为设计成可解释、可恢复、可审计。

Its SKILL.md is about 2.4k tokens, which your agent loads only when the skill is triggered. The skill folder holds 1 other file (for example `intent.json`).

The repository describes itself as: Intent-driven AI coding workflow runtime for consistent context, skills, approvals, validation, and handoffs across six AI coding hosts. The licence is AGPL-3.0.

Example prompts

  • “/ai-agent-system-architecture”

Workflow steps

5 steps, taken from the first numbered list in SKILL.md.

  1. 建立 Agent 行为图:用户输入、IntentSeedV1、目标、ContextReadPlanV2(V1 兼容)、Skill、工具、回执、确认、报告。
  2. 明确工具权限边界:宿主拥有文件、删除和命令权限,DevCodex 只约束工作流有效性并提供风险 advisory;adapter missing/failed/invalid/outside-workspace 不得建立本地 shadow deny;能力面发生变化时向中央…
  3. 定义上下文、记忆、handoff、summary 和恢复优先级。
  4. 设计状态机和失败恢复:blocked、retry、fallback、handoff。
  5. 用 direct replay、fixture replay、validate 或日志证据验证关键行为。

What it can do on your machine

Read from SKILL.md and the folder at commit 1dd4525. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md (its code samples are markdown).

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

AI Agent System Architecture loads about 2.4k tokens when it runs. Until then it costs about 35 tokens; SKILL.md has 572 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~35
When it runs · the whole SKILL.md, loaded when a task matches
~2.4k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from devcodex-labs/devcodex at commit 1dd4525, republished under its AGPL-3.0 licence (© devcodex-labs). 572 words, ~2,419 tokens.

Download SKILL.mdSave it as .claude/skills/ai-agent-system-architecture/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.
name
ai-agent-system-architecture
description
AI Agent 系统架构专家 Owner — 当任务涉及 Agent 路由、工具调用、上下文管理、记忆、状态机、权限、人机协作、可观测性、回放验证或模型辅助治理时使用;要求把 Agent 行为设计成可解释、可恢复、可审计。

AI Agent System Architecture Skill

定位

本 Skill 负责 AI Agent 系统 Owner 视角。它把 Agent 当作状态机和工具执行系统,而不是只关注 prompt 文案。

模型表达是非确定性输入,质量诊断与工具事实分别处理:内部标题、模板、阅读回执和短期入口元数据应能重建,不能成为整项任务硬门禁。恢复需保留原任务身份、实际宿主身份、精确目标与 operation 派发/效果记录;同一操作可读回后结算,未知副作用不得盲目重放。验证采用公共入口、表达变体、故障恢复和真实文件结果,不把固定文案断言当成原意遵守的证明。

触发条件

场景是否触发
Agent 路由、意图识别、Skill 触发、工具调用、权限、记忆、handoff、summary、hook 状态必须
任务涉及模型辅助治理、自动吸纳、自动复审、回放验证、可观测性或人机确认必须
报告或方案需要解释 Agent 为什么选择某工作流或工具必须
纯业务代码且不涉及 Agent 执行系统N/A + skipReason

核心门禁

Gate要求证据
AiAgentSystemArchitectureGateAgent 行为必须有路由、工具权限、上下文、状态机、观测和人机边界intentRouting、toolPermissionBoundary
AgentRoutingGate意图、Skill、模式和降级路径必须可解释intentRouting
ToolPermissionBoundaryGate工具权限、危险操作、确认和 fallback 必须明确toolPermissionBoundary
ContextAcquisitionGate每条消息必须先形成语义种子与唯一目标,再按计划读取最小充分上下文,并用 Post 成功回执证明完成IntentSeedV1、ContextReadPlanV2、ContextReadReceiptV2(V1 兼容)
ContextMemoryStateGate上下文恢复、记忆、handoff 和状态新鲜度必须设计contextMemoryModel
TaskContinuityWriteGate正式任务写入必须由单一 V5 store owner、精确 state/writer fence、canonical write context、operation 状态机与 terminal lineage 共同约束;TTL 不产生写权taskContinuityWriteContract
ReplayObservabilityGate行为验证不能只靠文字说明,需 replay、fixture 或日志证据observabilityReplay
TurnLivenessRecoveryGate长任务或工具输出后的 turn 必须用事件时间、AI-owned lease、continuation ACK、terminal invariant 与 checkpoint 区分运行、可疑、可恢复停滞和终态turnLivenessContract、TurnRecoveryCard、TurnLivenessEvidence
LocalTaskTraceGate当前 turn 的 typed trace 必须严格有序、拒绝重复/终态后追加,并只提供不执行 payload 的只读 replayLocalTaskTraceV1、LocalTaskTraceReplayV1
RepairCollaborationRoleBoundaryGaterepair task 必须把决策/验收与执行/验证角色、授权证据、状态与独立复证设计清楚;模型或 Agent 名称不构成风险分类roleAssignments、authorizationEvidence、independentReReview
AgentCapabilityDomainCompletenessGate声称完整/最终 Agent 架构或平台前先声明 completenessObject,并验证请求链、反馈链、横切面及适用产品/企业链agentCapabilityDomainMatrix
AgentCapabilityDomainCompletenessGate
completenessObject必查覆盖
kernelingress→cognition→context/knowledge→planning→execution/tools→response;observe→evaluate→evolve;governance/model/infrastructure
runtimekernel + local/hosted composition、state、tool registry、security、observability、replay
developer-productruntime + build→version→publish→deploy→invoke、SDK/CLI/API、local developer runtime、Agent Studio
hosted-platformdeveloper-product + run API、provider/connector/credential、deployment/fleet、tenant/workspace
enterprise-saashosted-platform + organization、entitlement、usage/metering/billing、admin/ops、audit/compliance

每个适用能力域必须记录 owner、publicPrivateBoundary、runtimeStatus、validationRoute。较窄对象通过不能升级解释为较宽对象完整;报告使用“完整/最终/无需新增域”时,缺少 completenessObject 或任一适用域即判 incomplete。

CapabilitySurfaceDecision 证据提供者

新增或改变 Agent 可调用能力面时,本 Skill 只向 spec-governance#CapabilitySurfaceDecisionGate 提供语义判断边界、model/application/user/host 控制方、read/write/execute 权限、状态机、authority、Task 协商、取消/超时/幂等和审计证据。它读取中央 decisionRef,不得自行决定或复制 preferredSurface 等 canonical 字段;MCP 能力、Tasks 或宿主行为没有 direct evidence 时保持 UNVERIFIED,不得由 Agent 名称或概念相似性推断支持。

ContextAcquisitionGate

Agent 上下文获取采用以下状态链,任何一步都不得用后一步的推断倒填:

text
IntentSeedV1 → unique project/activeRoot → ContextReadPlanV2(V1 兼容)→ attempted → PostToolUse observed → ContextReadReceiptV2(V1 兼容)→ project reality/final route
  • IntentSeedV1 仅来自当前消息语义和已观察到的 continuity,不得先全文读取 Profile / memory 再“识别”意图;关键词不是 canonical intent。
  • ContextReadPlanV2 必须显式区分 baseline、selected、excluded、unclassified 与 fullReadReason,并把稳定 planContentId 与 invocation planId 分离;ContextReadPlanV1 仅保留读取兼容。默认读取最小充分来源;Profile 规划阶段不得 hidden full read,记忆使用 bounded status/session/summary query。
  • ContextReadReceiptV2 只接受 planId、planContentId、contextEpoch、activeRoot、source identity/query 和结果精确关联的 PostToolUse 成功证据。PreToolUse、计算 cache hit、旧全文工具返回或 fallback 文案都不能声明 complete;V1 receipt 不具备跨 epoch delivery reuse 资格。
  • MCP 本地 stdio 正文观察先写入有界 ContextSourceObservationLedgerV1,lifecycle receipt 只作可重建投影;SkillRoute 仅在 epoch/plan/root/project 全同且 source metadata 新鲜时重放。旧快照覆盖不得丢失已交付证据,stale/blocked、source-digest 或 profile-drift 不得被 ledger 绕过。
  • SkillRoute per-turn envelope 是临时执行 cache,不是正式任务存储。容量必须按语义活跃对象计数;空 orphan、无业务义务终态和同会话已被后继 context 取代的未提交 route 只能在 root/turn lock、identity 与 quarantine/readback 复证后有界退出。protected、live lock、identity mismatch、其他会话未完成 route 和业务回复义务必须保留并在 hard pressure 下失败关闭。正式任务数量仍由 TaskRecoveryStoreV5 的字节/headroom 合同治理,不得套用 per-turn 数量上限。
  • 用户/项目明确要求、audit/migration、低置信或必要来源缺失可升级全量;目标、scope/action/risk、source digest 或 compact/resume 发生实质漂移时重新规划。不得每个动作都重复加载。
  • 宿主缺少结构化工具时可走一次 path-observable / instruction fallback;证据不足保持 partial/unverified,后续安全、CP、治理和验证门禁不得因节流而降低。
TaskContinuityWriteGate

正式任务的 locator、当前写者、单次 operation、终态和运行时投影是不同边界,不能把“找到了任务”解释为“可以写”。所有正式 envelope commit 必须由 TaskRecoveryStoreV5 精确比较 TaskRecoveryCommitFenceV1.stateSequence + writerGeneration;force 不绕过 stale fence,writer generation 只可在 owner transition 中增加 1。owner claim/transition 后必须 readback CanonicalTaskWriteContextV1,writer 绑定 task/root、lifecycle revision、state sequence、writer generation、holder session、operation/settled set digest、runtime generation 与 context digest。TTL 只作 freshness、清理和诊断,不能产生 takeover。

每个任务最多存在一个未结算 mutating TaskOperationRecordV1,固定经过 prepared → dispatched → observed → settled,未知副作用转 reconcile-required,未派发才允许 aborted-zero-effect。已派发操作不得自动重试;迟到回执和 emergency reserve 只能推进同一 operationId、idempotency key、writer generation、exact targets 与 before digest。terminal 必须同时验证 current writer、write context fence、settled set 和独立证据;replay 零新写,reopen 产生 lifecycle revision+1 与新 owner generation。

Show full SKILL.md (224 more words)Show less
TurnLivenessRecoveryGate

当任务涉及长时间运行、工具完成后无续接、线程持续 inProgress、恢复或宿主停滞时,先冻结 TurnLivenessContract:

字段要求
stateModel至少区分 idle / running / awaiting-continuation / suspect / stalled-recoverable / completed / error / interrupted
eventEvidenceturnKey / lastEventType / lastEventAt / lastToolOutputAt / continuationAckAt
lease只有已放行的 AI-owned operation 才能建立 lease;用户进程或未知 PID 不能作为可清理 lease
terminalInvarianttool output 不等于 turn 完成;显式 Stop/error/interruption 后必须清除 in-flight lease
checkpointphase / artifactPaths / nextAction / resumeToken / idempotencyKey,恢复前必须验证幂等边界
checkpointValidationresponse-time 与 post-execution 分开记录;缺 post evidence 只能 unverified/incomplete-timeout,实际 terminal evidence 才能 pass
localTaskTracetraceId/turnKey/status/sequence/openedAt/completedAt/events;eventId 唯一、sequence 从 1 递增、terminal 唯一且最后
capabilityBoundary分开记录 host-native watchdog、Hook event-time detection 与 read-only sidecar;Hook 无事件时不得宣称能自唤醒
validationdirect replay + no-continuation / active-lease / restart-rehydrate / duplicate-recovery fault matrix

默认 awaiting-continuation 可采用 120 秒 suspect / 300 秒 stalled advisory;慢模型推理和长工具必须由更长的 agent/operation lease 覆盖,不能机械套用 ACK 阈值。观察到 stale 只生成 TurnRecoveryCard;没有宿主授权与幂等复证时,禁止自动重放 mutation、kill/restart/interrupt/resume。

LocalTaskTraceV1 只保存当前 turn;历史 turn 由 TurnLiveness 摘要承接。重启时先校验 identity/sequence/duplicate/terminal,再生成 LocalTaskTraceReplayV1 数据投影;replay 的 stateMutation/operationReplay/payloadExecution/processControl 必须全部为 false。

执行步骤

  1. 建立 Agent 行为图:用户输入、IntentSeedV1、目标、ContextReadPlanV2(V1 兼容)、Skill、工具、回执、确认、报告。
  2. 明确工具权限边界:宿主拥有文件、删除和命令权限,DevCodex 只约束工作流有效性并提供风险 advisory;adapter missing/failed/invalid/outside-workspace 不得建立本地 shadow deny;能力面发生变化时向中央 decision 提供控制方、authority、状态与 Task 证据。
  3. 定义上下文、记忆、handoff、summary 和恢复优先级。
  4. 设计状态机和失败恢复:blocked、retry、fallback、handoff。
  5. 用 direct replay、fixture replay、validate 或日志证据验证关键行为。

输出字段

markdown
## AiAgentSystemArchitectureGate

| 字段 | 内容 |
|------|------|
| intentRouting | 意图、Skill、模式、降级路径 |
| toolPermissionBoundary | 工具权限、确认、危险操作和 fallback |
| contextMemoryModel | 上下文、记忆、handoff、summary、状态新鲜度 |
| contextAcquisition | seed、target、plan、selected/excluded、fullReadReason、fallback |
| contextReadReceipt | `ContextReadReceiptV2`(V1 兼容)的 Post 成功证据、内容身份、复用来源、缺失来源与完成状态 |
| stateMachineHandoff | 状态机、恢复、阻塞、交接 |
| observabilityReplay | replay、fixture、日志、validate 证据 |
| turnLivenessContract | 状态、事件、lease、ACK、终态、双阶段 checkpoint、LocalTaskTrace、能力边界与故障矩阵 |
| taskContinuityWriteContract | store owner、state/writer fence、CanonicalTaskWriteContext、operation 状态机、terminal/reopen、TTL 边界与负向探针 |
| humanInLoopBoundary | 用户确认、auto、人工复核和最终责任边界 |
| capabilitySurfaceEvidence | `decisionRef`、控制方、read/write/execute、authority、状态/Task 与直接证据边界 |
| evidenceMatrix | 判断 -> hook / runtime / report / memory / replay / tests |
| agentCapabilityDomainMatrix | completenessObject -> domain -> owner / boundary / runtime / validation |

反模式

反模式修正
把 Agent 能力写成 prompt 愿望写路由、状态机、工具权限和验证
用 summary 覆盖文件真相源遵循 Context Rehydration Contract
先全文读取 Profile / memory 再判断意图先形成 IntentSeedV1 与唯一目标,再执行有界计划和定向查询
把 PreToolUse、cache hit 或旧全文工具返回写成已完成仅以精确关联的 PostToolUse 成功结果生成 ContextReadReceiptV2;delivery reuse 还须同 session/epoch/source identity
自动模式绕过危险确认保留 S01/S06 等不可豁免底线
行为变更无 replay补 direct/fixture replay 或等价 validate
把 Hook 状态写入当成无事件 watchdog明确 event-time detection 边界,并用宿主能力或 gray read-only sidecar补充观察
trace replay 执行 payload 或恢复写操作只返回已校验的数据投影;写操作恢复必须另走授权和幂等复证

与其他 Skill 的关系

  • intent / routing:意图和 Skill 路由是 Agent 系统入口。
  • memory / summary:上下文恢复和状态新鲜度需要联动。
  • host-contract-verification:宿主行为变化需验证事件和可见回复。

© devcodex-labs, AGPL-3.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 1 other file in content/skills/ai-agent-system-architecture of devcodex-labs/devcodex.

  • SKILL.md
  • intent.json

Open the folder on GitHubat commit 1dd4525

Compare with similar skills

AI Agent System Architecture next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

AI Agent System Architecture compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
AI Agent System Architecture this skilldevcodex-labs/devcodex439—~2.4kAutomated safety check: PassAGPL-3.0
Agile Product Ownerdavila7/claude-code-templates32k2 repos~256Automated safety check: PassMIT
Agile Product Owneralirezarezvani/claude-skills28k3 repos~3.2kAutomated safety check: PassMIT
Super Product Ownersyahiidkamil/Software-Engineer-AI-Agent-Atlas401—~7.9kAutomated safety check: PassNone
Find Lead Account Ownerzapier/gtm-cheat-codes342—~275Automated safety check: PassMIT
Owner VoiceAnastasiyaW/codex-claude-code-config154—~874Automated safety check: NotesMIT

Similar skills

  • Agile Product Owner

    davila7/claude-code-templates

    Agile product ownership toolkit for Senior Product Owner including INVEST-compliant user story generation, sprint planning, backlog management, and velocity tracking.

    32k GitHub starsUsed in 2 repos~256 tokens
    Product & Project ManagementAuto-check passed
  • Agile Product Owner

    alirezarezvani/claude-skills

    Writes INVEST-checked user stories with acceptance criteria, splits epics, plans sprints from velocity and ranks the backlog with a weighted score.

    28k GitHub starsUsed in 3 repos~3.2k tokens
    Product & Project ManagementAuto-check passed
  • Super Product Owner

    syahiidkamil/Software-Engineer-AI-Agent-Atlas

    Complete Product Owner / Product Manager capability — a wiki-style knowledge map of product practice in active software development: the role and its boundaries, strategy cascade (vision → OKRs →…

    401 GitHub stars~7.9k tokensUpdated 3 mo ago
    Product & Project ManagementAuto-check passed
  • Find Lead Account Owner

    zapier/gtm-cheat-codes

    Official

    Resolve the right account owner for a lead using contact, company, and deal ownership evidence.

    342 GitHub stars~275 tokensUpdated 2 mo ago
    Productivity & AutomationAuto-check passed
  • Owner Voice

    AnastasiyaW/codex-claude-code-config

    Пишет и редактирует текст голосом владелицы — сообщение, письмо, пост, объяснение, урок, рабочий разбор, на русском и английском.

    154 GitHub stars~874 tokensUpdated yesterday
    Writing & ContentAuto-check: notes
  • Kanban Md Principal Owner

    antopolskiy/kanban-md

    Review kanban-md feature requests, issues, PRs, and design proposals for product fit, domain-model growth, configurability, and compatibility.

    223 GitHub stars~1.7k tokensUpdated 4 days ago
    Productivity & AutomationAuto-check passed

More from devcodex-labs/devcodex

All 70 skills in this repo
  • Accessibility I18n

    devcodex-labs/devcodex

    无障碍与国际化专家 Owner — 当任务涉及可访问性、键盘操作、焦点、屏幕阅读器、ARIA、语言地区、本地化、RTL、翻译资源、用户可见文案或多语言文档时使用;要求把包容性体验和本地化验证绑定到真实用户路径。

    439 GitHub stars~718 tokensUpdated 23 days ago
    Auto-check passed
  • API Contract Architecture

    devcodex-labs/devcodex

    API 契约架构专家 Owner — 当任务涉及 public API、HTTP/SDK/CLI 契约、版本兼容、错误模型、分页过滤、幂等、Schema、类型、迁移或消费者影响时使用;要求先冻结消费者契约,再设计实现与验证。

    439 GitHub stars~865 tokensUpdated 23 days ago
    Auto-check passed
  • Architecture Design

    devcodex-labs/devcodex

    架构设计文档编排 Owner — 当用户要求架构设计、系统设计、技术架构或可指导开发、Review 与任务拆分的完整方案时使用;要求从业务流程反推节点、状态、数据、一致性、异常补偿、ADR 与实施任务。

    439 GitHub stars~1.1k tokensUpdated 23 days ago
    Auto-check passed
  • Audit Common

    devcodex-labs/devcodex

    审查公共维度 G0~G5 + Profile Freshness Check — 所有 audit 子类型必先执行的基础维度层

    439 GitHub stars~4.1k tokensUpdated 23 days ago
    Auto-check passed
  • Audit Session

    devcodex-labs/devcodex

    审计工作流的跨会话状态机 — 在 <audit-root/.audit-state/<session-id.json 持久化轮次/发现项/收敛状态,支持 Token 中断后精准恢复

    439 GitHub stars~1.8k tokensUpdated 23 days ago
    Auto-check passed
  • Backend Domain Architecture

    devcodex-labs/devcodex

    后端领域架构专家 Owner — 当任务涉及领域模型、业务流程、权限、API、事务、一致性、幂等、兼容、数据边界、服务职责或用户要求从后端/领域专家角度审查时使用;要求用领域语言和业务不变量约束实现。

    439 GitHub stars~575 tokensUpdated 23 days ago
    Auto-check passed

Questions about AI Agent System Architecture

What does AI Agent System Architecture do?

AI Agent 系统架构专家 Owner — 当任务涉及 Agent 路由、工具调用、上下文管理、记忆、状态机、权限、人机协作、可观测性、回放验证或模型辅助治理时使用;要求把 Agent 行为设计成可解释、可恢复、可审计。. AI Agent System Architecture is an agent skill from devcodex-labs/devcodex.

How do I install AI Agent System Architecture in Claude Code?

Run `npx skills add devcodex-labs/devcodex --skill ai-agent-system-architecture -a claude-code`. Or copy the skill folder (content/skills/ai-agent-system-architecture in devcodex-labs/devcodex) into .claude/skills/ai-agent-system-architecture in your project. Claude Code loads it when a task matches its description.

How do I install AI Agent System Architecture in Codex?

Run `npx skills add devcodex-labs/devcodex --skill ai-agent-system-architecture -a codex`. Or copy the skill folder (content/skills/ai-agent-system-architecture in devcodex-labs/devcodex) into .agents/skills/ai-agent-system-architecture in your project. Codex loads it when a task matches its description.

Can I use AI Agent System Architecture in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add devcodex-labs/devcodex --skill ai-agent-system-architecture -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/ai-agent-system-architecture, .gemini/skills/ai-agent-system-architecture, .github/skills/ai-agent-system-architecture and .opencode/skills/ai-agent-system-architecture in your project.

What does AI Agent System Architecture need to run?

SKILL.md names no scripts, command-line tools or credentials: AI Agent System Architecture is instructions for the agent only.

Does AI Agent System Architecture access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is AI Agent System Architecture safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does AI Agent System Architecture use?

AI Agent System Architecture is published under the AGPL-3.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does AI Agent System Architecture use?

About 2.4k tokens (SKILL.md is roughly 9.7k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to AI Agent System Architecture?

Skills that share tags, products or a category with AI Agent System Architecture: Agile Product Owner (davila7/claude-code-templates, 32k stars), Agile Product Owner (alirezarezvani/claude-skills, 28k stars), Super Product Owner (syahiidkamil/Software-Engineer-AI-Agent-Atlas, 401 stars) and Find Lead Account Owner (zapier/gtm-cheat-codes, 342 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains AI Agent System Architecture?

devcodex-labs (a GitHub organization) maintains it in devcodex-labs/devcodex, which has 439 GitHub stars. The repository holds 70 skills in this directory. The repository was last updated on September 17, 2026.

Source: devcodex-labs/devcodex on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.