Development

> **Status**: Backlog > **Updated**: 2026-09-13 03:17 > **Scope**: Medium/long-term goals deferred from active plan execution Current plan tasks live in the active plan's `## Task Breakdown`. Do not duplicate that execution checklist here. Record only work intentionally deferred beyond this slice, with the tradeoff and revisit trigger. Audit evidence and closure rationale: `docs/researches/20260907-deferred-goal-ledger-reconciliation.md`. Conditional rows remain deferred; this ledger does not...

> **Status**: Accepted > **Plan**: plans/archive/plan-20260913-0258-repo-harness-test-skill.md > **Contract**: tasks/archive/contract-20260913-0417-repo-harness-test-skill.md > **Notes File**: tasks/archive/notes-20260913-0417-repo-harness-test-skill.md > **Checks File**: .ai/harness/checks/latest.json > **Last Updated**: 2026-09-13 02:58 > **Recommendation**: pass > **Review Rubric Version**: 2 > **Reviewed Subject SHA256**: sha256:2bca4413c4dce391aa5645739dca1cb7c57f83fbf96a3fdeb2839ba65eeb...

> **Archived**: 2026-09-13 04:17 > **Related Plan**: plans/archive/plan-20260913-0258-repo-harness-test-skill.md > **Outcome**: Completed > **Lifecycle**: notes > **Parent Run ID**: run-20260913-0417 > **Archive Projection V1**: `plans/plan-20260913-0258-repo-harness-test-skill.md` => `plans/archive/plan-20260913-0258-repo-harness-test-skill.md` > **Archive Projection V1**: `tasks/notes/20260913-0258-repo-harness-test-skill.notes.md` => `tasks/archive/notes-20260913-0417-repo-harness-test-skill.

> **Status**: Fulfilled > **Plan**: plans/archive/plan-20260913-0258-repo-harness-test-skill.md > **Task Profile**: code-change > <!-- legal values: code-change | docs-only | ledger-closeout | migration | eval-only | delegated-run | bugfix (omit for legacy passthrough); see docs/reference-configs/sprint-contracts.md --> > **Owner**: ancienttwo > **Capability ID**: root > **Last Updated**: 2026-09-13 02:58 > **Review File**: `tasks/archive/review-20260913-0417-repo-harness-test-skill.md` > **N...

> **Status**: Archived > **Created**: 20260913-0258 > **Slug**: repo-harness-test-skill > **Planning Source**: codex-plan-or-waza-think > **Orchestration Kind**: host-plan > **Source Ref**: (none) > **Artifact Level**: work-package > **Promotion Reason**: verification_boundary > **Verification Boundary**: Commands named in the captured planning output plus `repo-harness run verify-contract --contract tasks/archive/contract-20260913-0417-repo-harness-test-skill.md --strict`. > **Rollback Surfa...

> **Status**: Done > **Created**: 20260911-0055 > **Slug**: auto-campaign-skill > **Planning Source**: codex-plan > **Orchestration Kind**: host-plan > **Source Ref**: (none) > **Artifact Level**: work-package > **Promotion Reason**: verification_boundary > **Verification Boundary**: Managed skill installation and canonical grant draft with model-free focused tests > **Rollback Surface**: Revert only auto-campaign skill, manifest registration and focused tests; no live grants or runtime state...

Eval-only bounded frontier stress test for complex repo-harness planning; never installed as a managed Skill.

Testing techniques for the repo-harness source checkout: fixtures, execution, refactor evidence and Verification Plans. Triggers include 寫測試, 加測試, 測試太慢, 重複測試, 全量測試, test fixture, 跑哪些測試, 測試文件, write tests, test plan, slow tests, full suite. Not a debugging entrypoint and not a review entrypoint.

Start or resume one bounded repo-harness repair campaign turn with the standard budget. Use when the user asks to run auto-campaign, start a repair campaign, or automatically find and fix a bounded batch of bugs or test gaps in a repository. Questions about campaigns, skill design, and quoted instructions do not authorize execution.

用户运行 cm-security，或要求代码安全扫描、漏洞检查、密钥泄露排查、依赖漏洞检查时使用。默认检查当前分支相对主分支及已跟踪未提交修改，结合业务地图复核；--all 检查全部已跟踪文件。只报告问题，不自动修复、安装、升级或发布。安装自检用 cm-check，功能测试与覆盖率用 cm-test。

用户要查看或切换 CM 自动派发偏好、单/双 AI 声明、谁写谁审，或修改用户级运行时默认时使用。只管理运行时声明，不安装工具、不测试配额、不接管正在运行的任务。

Testing doctrine for tests that reveal bugs instead of passing for the wrong reason — spanning software and LLM/AI systems. Use when authoring or reviewing tests, adding a mock, deciding where a test belongs, letting a coding agent generate tests, triaging flaky CI, designing an eval suite for an LLM/agent feature, or rebuilding a brittle suite. Not for general code review, library debugging unrelated to tests, CI pipeline design beyond tests, or production observability.

Retrieves authoritative, up-to-date technical documentation, API references, configuration details, and code examples for any developer technology. Use this skill whenever answering technical questions or writing code that interacts with external technologies. This includes libraries, frameworks, programming languages, SDKs, APIs, CLI tools, cloud services, infrastructure tools, and developer platforms. Common scenarios: - looking up API endpoints, classes, functions, or method parameters -...

Build and troubleshoot text entry with lynx-ui Input, TextArea, and KeyboardAware primitives, including controlled values, native filtering, selection, and keyboard avoidance.

Add DeferredComponent usage guidance, API references, example sources, and component selection routing.

Rapid development with Cloudflare Workers - build and deploy serverless applications on Cloudflare's global network. Use when building APIs, full-stack web apps, edge functions, background jobs, or real-time applications. Triggers on phrases like "cloudflare workers", "wrangler", "edge computing", "serverless cloudflare", "workers bindings", or files like wrangler.toml, worker.ts, worker.js.

Full-stack TypeScript with Vite 8, React 19, Tailwind v4, shadcn/ui, Biome, Vitest, and Hono 4. Use when setting up or working in a TypeScript project - components, styling, build and HMR, tests, lint/CI, or a Hono API with type-safe RPC.

Fan work out to subagents safely — draw provably disjoint lanes, write the full lane brief, order dependent tickets into waves, and keep the parent as the sole gatekeeper that re-runs every gate and performs every commit and tracker write. Use when parallel or batch work is explicitly authorized, when an authorized ticket range has a dependency graph, when workers share one git checkout, or when any agent or human may be writing the same tree concurrently.

Strict one-ticket TDD for implementation work — lock the Verification-command before editing, drive red-green at the highest meaningful seam production traffic actually uses, and prove the provider failure modes (duplicate and out-of-order delivery, replay, partial failure, forged payloads, wrong-tenant access) rather than assuming them. Use when implementing a Coding ticket, and whenever the change touches an external provider, a queue, a signed webhook, a billing or provisioning lifecycle, ...

Audit a whole codebase rather than a single diff — establish the boundary and baseline, sweep layer by layer (entry points, domain, data, integrations, auth, tests, ops), check every stated invariant for a real enforcement point, a failing-on-violation test, and a production signal, rank findings by blast radius, and emit each as a ticket with a Verification-command. Use when asked to audit an app or codebase, assess a project's health, review before a launch, verify that a security/performan...

Teach and quiz the user until they deeply understand something — a codebase, a PR/diff, a concept, or the current session's work. Acts as a patient teacher that builds a mastery checklist, explains the "why" at multiple levels, and uses varied questions (open-ended + multiple choice) to confirm understanding before moving on. Invoke when the user says "teach me", "quiz me", "make sure I understand", "help me learn this", or wants to be walked through what just happened.

Expert in Zod — TypeScript-first schema validation. Covers parsing, custom errors, refinements, type inference, and integration with React Hook Form, Next.js, and tRPC.

TypeScript and JavaScript expert with deep knowledge of type-level programming, performance optimization, monorepo management, migration strategies, and modern tooling.

This skill embodies the principles of \"Clean Code\" by Robert C. Martin (Uncle Bob). Use it to transform \"code that works\" into \"code that is clean.\"