跳到正文
cccskills
免费在 GitHub 公开

geode-workflow

GEODE evidence-first development workflow. Use for feature work, provider/model changes, GUI/computer-use work, PDF/document ingestion, observability/schema/event/state changes, verification planning, GitFlow merge prep, and requests to "follow the workflow" or "scaffold the process".

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(7)

  • SKILL.md3.6 KB
  • references/codex-geode-paired-coding.md6.2 KB
  • references/gitflow.md1.4 KB
  • references/observability-contract.md5.0 KB
  • references/phase-checklist.md3.6 KB
  • references/provider-grounding.md3.6 KB
  • references/verification-gates.md11.5 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

GEODE Workflow

Use this skill as GEODE's default execution scaffold. It replaces the older issue-first workflow with an evidence-first loop: inspect existing code, ground external contracts, define the observable surface, implement narrowly, verify, then report honestly.

Core Loop

Follow the canonical workflow and its execution scope. Produce the requested reviewable result; load only the references relevant to the task. For abstraction, naming, type/class, schema, test, site, or version decisions, use $geode-code-conventions.

Reference Routing

Read only the reference needed for the current task:

NeedRead
Phase-by-phase checklist for ordinary code workreferences/phase-checklist.md
Abstraction, naming, type/class, schema, test, site, and version conventions$geode-code-conventions (follow its task-specific SOT routing)
Provider/model/API capability claimsreferences/provider-grounding.md
Schema/log/event/state/trajectory consistencyreferences/observability-contract.md
Usage/cache fields, estimates versus billing, incomplete accountingUsage accounting contract
Same-task Codex–GEODE production and comparisonreferences/codex-geode-paired-coding.md
Test, lint, type, prompt, full-suite reporting, or failure-to-rule reviewreferences/verification-gates.md
Branch, PR, merge, and cleanup operationsreferences/gitflow.md

For simple docs or trivial bug fixes, use the core loop and skip references that do not affect the task.

Non-Negotiables

  • Do not switch branches inside an active worktree.
  • Do not revert unrelated dirty work.
  • Architecture/extensibility implementation work starts only after a roadmap-only claim PR moves the package from READY to IN_PROGRESS on canonical develop and names its owner/implementation branch. Claim and reconciliation PRs establish that prerequisite; they do not require an earlier claim. The implementation PR preserves the claimed status and never predicts IN_DEVELOP or DONE.
  • Register architecture scope discovered during implementation through a separate roadmap-only GAP-registration PR. Registration may add an OPEN package without a prior claim, but does not authorize implementing it.
  • Within the architecture/extensibility program, use origin/main only for a tracking-only roadmap DONE worktree after release. Its PR targets main, carries no implementation code. The next owned feature branch incorporates the resulting main history before CI and integration; no standalone sync PR.
  • Do not mark unsupported provider behaviour as supported without grounding.
  • Do not add ad hoc observability strings when the capability/evidence vocabulary should be extended.
  • Do not log raw screenshots, base64 blobs, API keys, tokens, passwords, or secret fragments.
  • Do not imply a full-suite pass if only targeted checks ran.

Output Contract

When finishing, include:

  • implementation summary
  • key files changed
  • verification commands and results
  • known failures, skipped gates, or live-test uncertainty
  • GitFlow state if any branch/PR/merge action was requested

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Audit agent runtimes and their tests for lifecycle inference, capability overload, context pollution, correlated verification, prompt-only control, false success, dead compatibility paths, test-only seams, and complexity without contribution. Use for whole-repository agent audits, slop or dead-code triage, test contribution reviews, large refactors, tool/context simplification, and cleanup requests that need proof before deleting code or tests.

中文简介正在准备中。当前显示原文说明。

mangowhoiscloud/geode152026年10月16日 更新

Diagnose agent operational failures through default-policy checks, multi-gap analysis, explicit service ownership, request-context lifetime, and degradation-versus-correctness checks.

中文简介正在准备中。当前显示原文说明。

mangowhoiscloud/geode152026年10月16日 更新

Align, run, audit, and publish GEODE benchmark comparisons against Agent-World arXiv 2604.18292v1 across MCP-Mark, BFCL V4, and tau2-bench. Use for Agent-World or AgentWorld comparison requests, cross-suite agent benchmark design, Tau2 comparator/profile diagnosis, paired thin-runtime-versus-GEODE experiments, mean-accuracy@8 replication, or benchmark comparability and artifact reviews.

中文简介正在准备中。当前显示原文说明。

mangowhoiscloud/geode152026年10月16日 更新

Verification checklist to prevent fake success. Detects test deletion/disabling, coverage regression, lint bypass, secret exposure. Triggered by "deception", "fake" ("가짜"), "fake success", "verification" ("검증"), "checklist" ("체크리스트"), "deletion detection" ("삭제 탐지"), "regression" keywords.

中文简介正在准备中。当前显示原文说明。

mangowhoiscloud/geode152026年10月16日 更新

Implement proven backend architecture patterns including Clean Architecture, Hexagonal Architecture, and Domain-Driven Design. Use when architecting complex backend systems or refactoring existing applications for better maintainability.

中文简介正在准备中。当前显示原文说明。

mangowhoiscloud/geode152026年10月16日 更新

Auto-search and summarize latest AI/agent papers.

中文简介正在准备中。当前显示原文说明。

mangowhoiscloud/geode152026年10月16日 更新

查看 mangowhoiscloud 的全部技能

报告此技能的问题