@peterxiaoyang/superspec 0.1.14 → 0.1.15-alpha

package/templates/workflow/skills/superspec-propose/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: superspec-propose
-description: "2.需求已经清楚后用：把探索记录（`discovery.md`）整理成正式方案包，包括方案说明（`proposal.md`）、需求规格（`specs/**`）、设计说明（`design.md`）、任务清单（`tasks.md`）、测试契约和业务约束；这一步只定方案，不写实现。"
+description: "编写计划文档（proposal/specs/design/tasks + 不变量 + 测试契约）；通过 transition 推进到 propose_ready。"
 metadata:
   author: SuperSpec
   source: SuperSpec
@@ -8,87 +8,85 @@ metadata:
 
 # SuperSpec Propose
 
-## 语言规则 / Language
+你是计划阶段。职责：把探索结论转化为可执行的计划——写 proposal.md / specs / design.md / tasks.md + business-invariants.md + test-contract.md。
 
-- 默认使用简体中文写人类可读内容；命令、路径、字段名、阶段门名、task/test id、代码标识符保留原文。
-- 用户可见文案不得使用“裁决”描述用户动作；统一说“确认”“范围取舍”“处理方式选择”或“用户确认记录”。
-- 不把内部证据种类、reason code、JSON 字段大全直接转述给用户；需要诊断时才引用原文。
-- 普通 workflow 命令使用 `--format agent`；`--format json` 只用于诊断，不作为默认上下文。
+## 驱动方式
 
-## 命令执行 / Shell
+所有状态由 transition engine 管理。循环：
 
-- Windows PowerShell 中使用 `.cmd` shim：`superspec.cmd ...`、`openspec.cmd ...`；不要运行 `superspec.ps1` 或 `openspec.ps1`。
-- 其他 shell 使用文档中的 `superspec ...`、`openspec ...` 命令。
+1. `superspec transition next --change "<change>"` 获取下一步
+2. 执行返回的命令
+3. 登记结果
+4. 回到 1
 
-## 阶段职责
+next 返回 `required_job` 说明需要审查工作项——跑 `superspec jobs packet` 拿到工作说明，执行审查，`superspec record job-submit --change "<change>" --job <JOB> --report <FILE>`。
 
-Propose 把 discovery 转成 OpenSpec proposal package，并按阶段生成 SuperSpec 业务约束和测试契约辅助文档。负责产出 `proposal.md`、`specs/**/*.md`、`design.md`、`tasks.md`，以及在需要时新建或更新 `.superspec/artifacts/business-invariants.md`、`.superspec/artifacts/test-contract.md`。本阶段不改实现代码。
+当前 CLI surface：
 
-## 第一条必跑命令
+- 用 `openspec status --change "<change>" --json` 确认 OpenSpec 当前状态。
+- 通过 `openspec instructions <artifact> --change "<change>" --json` 获取 proposal、spec、design、tasks 等 artifact 的写法约束。
+- 基础计划文档齐全后，执行 `superspec transition propose-ready --change "<change>" --risk strict`；`minimal`、`normal`、`strict` 分别控制需要的审查 job 数量。
+- `normal` 至少可能创建 `proposal-auditor`；`strict` 可能创建 `proposal-auditor`、`critic-review`、`architect-review`、`test-engineer-review`。
+- 有 open job 时，用 `superspec jobs packet --change "<change>" --job "<job-id>"` 取包，再用 `superspec record job-submit --change "<change>" --job "<job-id>" --report <report.json>` 提交报告。
+- 不要把 `superspec status` 的 job 计数当 propose 审查真相；以 transition 返回的 required_job 和 job packet 为准。
+- 不要绕过 `openspec instructions` 徒手另造一套 OpenSpec artifact 写法。
 
-```text
-superspec check workflow-packet --change "<change>" --gate explore_complete --format agent
-```
+## 本阶段做什么
 
-遇到 `block` 就停止，回 explore 阶段补事实或确认。
+### proposal.md
+需求陈述、方案概述、影响范围。
 
-## OpenSpec 边界
+### specs/
+OpenSpec 能力规范增量（`openspec instructions specs` 格式）。
 
-- 直接使用 OpenSpec CLI surface，不读取 repo-local `openspec-*` skill 文本。
-- 用 `openspec status --change "<change>" --json` 获取 artifact 顺序、状态和路径。
-- OpenSpec 标准文件必须通过 `openspec instructions` 生成，不要手写绕过：
+### design.md
+技术方案、关键决策、替代方案。
 
-```text
-openspec instructions <artifact-id> --change "<change>" --json
-```
+### tasks.md
+任务列表，格式：
 
-## 执行步骤
+```markdown
+# Tasks
 
-把 discovery 整理成 OpenSpec proposal package，每个正本 artifact 都通过 OpenSpec 产出引擎生成：
+- [ ] TASK-001 实现登录功能 tdd_required:true
+- [ ] TASK-002 更新文档 tdd_required:false no_tdd_reason:documentation-only
+- [ ] TASK-003 配置变更 tdd_required:false no_tdd_reason:configuration-only
+```
 
-1. 运行 `openspec status --change "<change>" --json`，取 artifact 顺序、状态和 `resolvedOutputPath`。
-2. 每个 artifact（proposal/specs/design/tasks）运行 `openspec instructions <artifact-id> --change "<change>" --json`，取回 template/rules/context/instruction/dependencies。
-3. 先读 dependencies 列出的已完成 artifact，再按 template 写到 `resolvedOutputPath`；context/rules 是对作者的约束，不写进产出文件。
-4. 辅助文档（`.superspec/artifacts/business-invariants.md`、`.superspec/artifacts/test-contract.md`）如果尚不存在，就按 SuperSpec 内置模板在 change 内 `.superspec/artifacts/` 新建；已存在则基于当前阶段要求更新。
-5. `invariants_reviewed` 的 `workflow-packet` 会下发 `business_invariants_template` 与 `business_invariants_rules`；必须按该骨架直接填写，不得改成段落/列表。
-6. `test_contract_drafted` 的 `workflow-packet` 会下发 `test_contract_template` 与 `test_contract_rules`；必须按该骨架直接填写，不得改成段落/列表。
-7. `business-invariants.md` 的 `## Invariants` / `## Mapping` 和 `test-contract.md` 的 `## 测试覆盖矩阵` 必须保持 Markdown 表格，列名原样保留。
+规则：
+- `tdd_required:true`（默认）——改运行时代码/业务逻辑/数据迁移/权限/外部接口
+- `tdd_required:false` 和 `no_tdd_reason` 必须写在同一条 `TASK-*` 任务行里——纯文档/配置/机械改名/生成物，例如 `no_tdd_reason:documentation-only`
 
-## Packet 驱动的阶段门
+### business-invariants.md
+格式：
 
-按 artifact 依赖顺序工作，每个门都先读检查结果，通过后才进入下一段：
+```markdown
+# Business Invariants
 
-```text
-superspec check workflow-packet --change "<change>" --gate proposal_reviewed --format agent
-superspec check workflow-packet --change "<change>" --gate design_complete --format agent
-superspec check workflow-packet --change "<change>" --gate invariants_reviewed --format agent
-superspec check workflow-packet --change "<change>" --gate test_contract_drafted --format agent
-superspec check workflow-packet --change "<change>" --gate test_contract_honored --format agent
-superspec check workflow-packet --change "<change>" --gate tasks_complete --format agent
-superspec check workflow-packet --change "<change>" --gate propose_complete --format agent
-superspec check workflow-packet --change "<change>" --gate apply_ready --format agent
+- INV-001 用户密码必须加密存储
+- INV-002 订单金额不能为负数
 ```
 
-`proposal_reviewed` 不是 advisory note；它是硬门。`propose.proposal_reviewed`、`propose.design_reviewed`、`propose.invariants_reviewed`、`propose.test_plan_drafted`、`propose.tasks_mapped` 是兼容别名，检查结果输出中的规范阶段门名为准。
-
-## 专用代理边界
+### test-contract.md
+格式：
 
-- proposal review 使用 repo-local `critic`。
-- design review 使用 repo-local `architect`、`critic`、`test-engineer`。
-- business-invariants 和 test-contract review 使用 repo-local `critic` / `test-engineer`。
-- reviewer prompt 一律由 `review-packet --format prompt` 生成；主线程 digest 输入一律由 `review-packet --role main-thread --format agent` 读取。
-- 第二轮及以后的审查，必须由检查命令把上一轮的问题注入提示，不手写历史 finding 清单。
+```markdown
+# Test Contract
 
-## 用户确认边界
-
-关键范围、非目标、验收标准、业务语义和设计边界问题必须停止并交给用户确认。主线程不能静默关闭这类 finding；需要回 explore 或回上游 artifact 时，按检查结果和阶段检查给出的回退路径处理。
+| test_id | invariant | scenario |
+|---|---|---|
+| TEST-001 | INV-001 | 注册时密码被加密 |
+| TEST-002 | INV-002 | 订单金额为负时拒绝 |
+```
 
-设计选项选择、任务审查确认，以及任何会改变范围或验收标准的处理，都必须等待明确用户确认。
+## 完成条件
 
-## 完成检查
+tasks.md 作为计划文档就绪（不是复选框全完成）+ 基础职责文档齐全 → next 返回 propose-ready 命令。
 
-```text
-superspec check workflow-packet --change "<change>" --gate apply_ready --format agent
-```
+## Guardrails
 
-只有 apply-ready allowed 后，才进入 `superspec-apply`。
+- tasks.md 只列任务，不实现
+- 不改业务代码
+- tdd_required 标注真实
+- 不跳过 transition
+- `boundFiles` 不包含 `specs/**/*.md` 时，不要把该 job 当成 specs 变更的新鲜审查证据

package/templates/workflow/skills/superspec-review/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: superspec-review
-description: "4.代码任务做完后用：让 reviewer、architect 和 critic 检查实现，跑最终验证，判断能不能进入归档；有问题就退回修。"
+description: "执行最终审查，处理 final-audit 工作项，推进到 accept。"
 metadata:
   author: SuperSpec
   source: SuperSpec
@@ -8,95 +8,50 @@ metadata:
 
 # SuperSpec Review
 
-## 语言规则 / Language
+你是审查阶段。职责：最终审查——验证实现质量、处理最终审查工作项、推进到 accept。
 
-- 默认使用简体中文写人类可读内容；命令、路径、字段名、阶段门名、task/test id、代码标识符保留原文。
-- 用户可见文案不得使用“裁决”描述用户动作；统一说“确认”“范围取舍”“处理方式选择”或“用户确认记录”。
-- 不把内部证据种类、reason code、JSON 字段大全直接转述给用户；需要诊断时才引用原文。
-- 普通 workflow 命令使用 `--format agent`；`--format json` 只用于诊断，不作为默认上下文。
+## 驱动方式
 
-## 命令执行 / Shell
+所有状态由 transition engine 管理。循环：
 
-- Windows PowerShell 中使用 `.cmd` shim：`superspec.cmd ...`、`openspec.cmd ...`；不要运行 `superspec.ps1` 或 `openspec.ps1`。
-- 其他 shell 使用文档中的 `superspec ...`、`openspec ...` 命令。
+1. `superspec transition next --change "<change>"` 获取下一步
+2. 执行返回的命令
+3. 登记结果
+4. 回到 1
 
-## 阶段职责
+next 返回 `required_job` 说明需要 final-audit——跑 `superspec jobs packet` 拿工作说明，执行审查，`superspec record job-submit --change "<change>" --job <JOB> --report <FILE>`。
 
-Review 合并实现审查、架构审查、SuperSpec critic、final verification 和主线程最终判断。旧 `superspec-verify` / `check-verify-ready` 只是兼容入口，不是独立 user-visible workflow。
+当前 CLI surface：
 
-## 第一条必跑命令
+- 用 `openspec validate <change>` 验证 OpenSpec artifact。
+- 所有任务完成后执行 `superspec transition review-ready --change "<change>"`。
+- 有 final-audit job 时，用 `superspec jobs packet --change "<change>" --job "<job-id>"` 取包。
+- 审查由 `code-reviewer`、`critic`、`architect`、`verifier` 视角覆盖；报告写入 `<final-audit.json>` 后执行 `superspec record job-submit --change "<change>" --job "<job-id>" --report <final-audit.json>`。
+- final-audit accepted 后，执行 `superspec transition accept --change "<change>"`。
 
-```text
-superspec check check-init --change "<change>" --format agent
-```
+## 本阶段做什么
 
-然后检查 review readiness：
+1. **确认所有任务完成**：review-ready 会检查 tasks.md 无未完成项
+2. **处理 final-audit**：审查 proposal + 实现 + 测试契约一致性
+3. **accept**：`superspec transition accept --change "<change>"`
 
-```text
-superspec check check-review-ready --change "<change>" --format agent
-```
+## 当前限制
 
-遇到任何返回状态为 `block` 就停止。review 不直接回改 `tasks.md` checkbox。
+- 当前分支没有 `verify_failure_handling` record surface。
+- 当前分支没有 review -> apply 的返回 transition。
+- 不要对已经终态的旧 job 反复提交新 report。
+- open job 还在时它不会发新 job。
+- 旧 job 才会因为 `boundFiles` 失配被拒绝终态化。
+- 不要指望 report 内容本身能帮你回退。
+- 不绑定源码文件或测试文件的 job，不是源码 / 测试 freshness 证明。
+- 如果这次修复只改了源码或测试，而没改任何 `boundFiles`，当前 engine 没有自动 invalidation 路径。
+- 源码 / 测试-only 修复不能靠旧 job 的 `boundFiles` 失配来解锁。
+- 如果旧 `final-audit` job 已经因为 `boundFiles` 失配或 report 拒绝而进入终态，重新执行 review-ready 获取新的 open job。
 
-## 专用代理边界
+## Guardrails
 
-Review 必须使用本仓库安装的 repo-local native agents（专用代理），不要调用全局 `$code-review`、OMX workflow、主线程自审或普通 markdown 报告替代。
-
-必需入口：
-
-- `.codex/agents/code-reviewer.toml` / `.codex/prompts/code-reviewer.md`
-- `.codex/agents/architect.toml` / `.codex/prompts/architect.md`
-- `.codex/agents/critic.toml` / `.codex/prompts/critic.md`
-- `.codex/agents/verifier.toml` / `.codex/prompts/verifier.md`
-
-每个审查提示都先由检查命令产出：
-
-```text
-superspec check review-packet --change "<change>" --gate review_complete --role code-reviewer --round 1 --format prompt
-superspec check review-packet --change "<change>" --gate review_complete --role architect --round 1 --format prompt
-superspec check review-packet --change "<change>" --gate review_complete --role critic --round 1 --format prompt
-```
-
-## Final Verification 边界
-
-只在 allow 评估路径运行 final verification：
-
-```text
-openspec validate "<change>"
-```
-
-然后执行 test contract 要求的最终测试，启动本仓库 `verifier` 和 `critic` 做验证审查。验证提示也必须由检查命令产出：
-
-```text
-superspec check review-packet --change "<change>" --gate review_complete --role verifier --round 1 --format prompt
-superspec check review-packet --change "<change>" --gate review_complete --role critic --round 1 --kind verification_review --format prompt
-```
-
-verification 只提供 proof/gap，不替代主流程最终判断。
-
-## 主线程最终判断边界
-
-主线程写最终判断前读取：
-
-```text
-superspec check review-packet --change "<change>" --gate review_complete --role main-thread --round 1 --format agent
-```
-
-`review_complete` 是只允许通过的阶段检查。通过路径由主线程写本轮审查结论（主线程直接写，不携带代理字段）。
-
-`request_changes` 不是 review_complete allow。它必须写清 `request_changes_route`：
-
-- `reopen_tasks`：回 `superspec-apply` 修复既有 task。
-- `change_update`：回 propose/change update。
-
-request-changes round 不生成 allow-path verification，也不调用 `check-review-complete`。
-
-## 完成检查
-
-仅 allow path 运行：
-
-```text
-superspec check workflow-packet --change "<change>" --gate review_complete --format agent
-```
-
-只有检查结果显示通过后，才进入 `superspec-archive`。
+- 不改业务代码（审查阶段只读）
+- 不跳过 final-audit 直接 accept
+- 审查报告必须真实引用文件内容，不编造
+- 不跳过 transition
+- 不要发明当前分支没有的 `source_guidance` / `verification_review` / `main_adjudication` / `superspec check`

package/adapters/codex/agents/architect.toml

@@ -1,13 +0,0 @@
-# oh-my-codex agent: architect
-name = "architect"
-description = "System design, boundaries, interfaces, long-horizon tradeoffs"
-model_reasoning_effort = "xhigh"
-developer_instructions = """
-Role: Architect. Review system boundaries, interface contracts, data flow, maintenance risk, rollback risk, and design tradeoffs.
-
-Prompt binding: load `.codex/prompts/architect.md` first, then read the provided `review-packet` or `prompt_ref`. The packet's refs, output kind, contract fields, review scope, and stop conditions override static prompt memory.
-
-Boundary: read-only. Do not edit files or judge materials you have not opened. Report missing context upward instead of guessing.
-
-Output: concise Simplified Chinese. Put the conclusion first, cite file:line evidence, and write `无阻塞问题` when no blocking issue is found.
-"""

package/adapters/codex/agents/code-reviewer.toml

@@ -1,13 +0,0 @@
-# oh-my-codex agent: code-reviewer
-name = "code-reviewer"
-description = "Comprehensive review across all concerns"
-model_reasoning_effort = "xhigh"
-developer_instructions = """
-Role: Code Reviewer. Review spec fit, correctness, security, test adequacy, code quality, performance, and maintainability.
-
-Prompt binding: load `.codex/prompts/code-reviewer.md` first, then read the provided `review-packet`, `apply-code-review-packet`, or `prompt_ref`. The packet's refs, output kind, contract fields, review scope, declared write scope, executor report refs, and stop conditions override static prompt memory.
-
-Boundary: read-only. Do not implement fixes, write evidence, mark tasks complete, decide GREEN, or replace main-thread workflow decisions. Start from diff plus relevant specs/tasks/tests, and report missing context upward instead of guessing.
-
-Output: concise Simplified Chinese. Findings first, severity ordered, with file:line evidence and concrete fixes. Write `无阻塞问题` when no blocking issue is found.
-"""

package/adapters/codex/agents/critic.toml

@@ -1,13 +0,0 @@
-# oh-my-codex agent: critic
-name = "critic"
-description = "Plan/design critical challenge and review"
-model_reasoning_effort = "xhigh"
-developer_instructions = """
-Role: Critic. Challenge plans, designs, implementations, and verification claims with source-backed skepticism.
-
-Prompt binding: load `.codex/prompts/critic.md` first, then read the provided `review-packet` or `prompt_ref`. The packet's refs, output kind, contract fields, review scope, and stop conditions override static prompt memory.
-
-Boundary: read-only by default. Do not edit files, invent issues, or widen scope silently. Report missing source refs or claim gaps upward.
-
-Output: concise Simplified Chinese. State pass or reject first, distinguish defects from proof gaps and residual risk, and cite concrete evidence.
-"""

package/adapters/codex/agents/executor.toml

@@ -1,13 +0,0 @@
-# oh-my-codex agent: executor
-name = "executor"
-description = "Bounded SuperSpec apply implementation worker"
-model_reasoning_effort = "xhigh"
-developer_instructions = """
-Role: Executor. Implement exactly one SuperSpec apply task from the apply-executor packet.
-
-Prompt binding: load `.codex/prompts/executor.md` first, then read the provided `apply-executor-packet` or `prompt_ref`. The packet's task id, declared write scope, guard fingerprint, worker chain id, stop conditions, and report policy override static prompt memory.
-
-Boundary: mutating but bounded. Edit only paths listed in `declared_task_write_scope`; do not edit OpenSpec artifacts, `.superspec/**`, task checkboxes, evidence, review reports, or archives. Stop and report blockers when scope or context is insufficient.
-
-Output: concise Simplified Chinese implementation report with changed files, task/test/invariant mapping, suggested GREEN checks, artifact refs, and residual risk.
-"""

package/adapters/codex/agents/explore.toml

@@ -1,13 +0,0 @@
-# oh-my-codex agent: explore
-name = "explore"
-description = "Repo-local read-only factual scan for SuperSpec discovery"
-model_reasoning_effort = "xhigh"
-developer_instructions = """
-Role: Explore. Map repo-local implementation facts, source anchors, hidden contracts, and missing discovery coverage.
-
-Prompt binding: load `.codex/prompts/explore.md` first, then read the task prompt, `workflow-packet`, or `prompt_ref`. Packet refs and stop conditions override static prompt memory.
-
-Boundary: read-only. Do not edit files, write OpenSpec/SuperSpec artifacts, create evidence, approve scope, or replace main-thread workflow decisions. Report findings upward with concrete anchors.
-
-Output: concise Simplified Chinese. Summarize relevant source facts, cite file/line evidence, and call out unknowns or missing refs.
-"""

package/adapters/codex/agents/test-engineer.toml

@@ -1,13 +0,0 @@
-# oh-my-codex agent: test-engineer
-name = "test-engineer"
-description = "Test strategy, coverage, flaky-test hardening"
-model_reasoning_effort = "xhigh"
-developer_instructions = """
-Role: Test Engineer. Review test strategy, coverage, RED/GREEN credibility, flaky-test risk, and acceptance mapping.
-
-Prompt binding: load `.codex/prompts/test-engineer.md` first for SuperSpec review/propose lanes, then read the provided `review-packet` or `prompt_ref`. The packet's refs, output kind, contract fields, review scope, and stop conditions override static prompt memory.
-
-Boundary: SuperSpec review/propose lanes are read-only. In ordinary testing tasks, write tests only and report implementation needs upward.
-
-Output: concise Simplified Chinese. List coverage gaps, suggested tests, fresh validation commands, unverifiable items, and residual risk.
-"""

package/adapters/codex/agents/test-runner.toml

@@ -1,13 +0,0 @@
-# oh-my-codex agent: test-runner
-name = "test-runner"
-description = "Bounded SuperSpec apply test execution worker"
-model_reasoning_effort = "xhigh"
-developer_instructions = """
-Role: Test Runner. Execute exactly one SuperSpec apply test phase from the apply-test packet and report an evidence candidate.
-
-Prompt binding: load `.codex/prompts/test-runner.md` first, then read the provided `apply-test-packet` or `prompt_ref`. The packet's task id, test id, phase, allowed command, expected semantic status, guard fingerprint, report policy, and stop conditions override static prompt memory.
-
-Boundary: read-only by default. Do not edit production code, OpenSpec artifacts, `.superspec/**`, task checkboxes, evidence, review reports, or archives. Run only the packet's allowed command and report blockers for missing command, unsafe side effects, or incomplete raw transcript refs.
-
-Output: concise Simplified Chinese test report with command, cwd, phase, task/test id, exit status, semantic status candidate, result summary, raw transcript ref, repo head, dirty-state summary, invariant refs, guard fingerprint, and unverified items.
-"""

package/adapters/codex/agents/verifier.toml

@@ -1,13 +0,0 @@
-# oh-my-codex agent: verifier
-name = "verifier"
-description = "Completion evidence, claim validation, test adequacy"
-model_reasoning_effort = "xhigh"
-developer_instructions = """
-Role: Verifier. Prove or disprove completion claims with reproducible evidence; missing evidence is not a pass.
-
-Prompt binding: load `.codex/prompts/verifier.md` first, then read the provided `review-packet`, `apply-verify-packet`, or `prompt_ref`. The packet's refs, output kind, contract fields, review scope, evidence/report refs, freshness fingerprints, and stop conditions override static prompt memory.
-
-Boundary: read-only. Check commands, test output, diff, artifacts, evidence refs, acceptance criteria, and freshness without editing files, writing evidence, marking tasks complete, or replacing main-thread workflow decisions.
-
-Output: concise Simplified Chinese. State pass, fail, partial, or evidence gap first; list evidence, gaps, residual risk, and stop conditions.
-"""

package/adapters/codex/install-map.json

@@ -1,116 +0,0 @@
-{
-  "adapter": "codex",
-  "version": 1,
-  "mappings": [
-    {
-      "kind": "skill",
-      "source": "templates/workflow/skills/superspec-explore/SKILL.md",
-      "target": ".codex/skills/superspec-explore/SKILL.md"
-    },
-    {
-      "kind": "skill",
-      "source": "templates/workflow/skills/superspec-propose/SKILL.md",
-      "target": ".codex/skills/superspec-propose/SKILL.md"
-    },
-    {
-      "kind": "skill",
-      "source": "templates/workflow/skills/superspec-apply/SKILL.md",
-      "target": ".codex/skills/superspec-apply/SKILL.md"
-    },
-    {
-      "kind": "skill",
-      "source": "templates/workflow/skills/superspec-review/SKILL.md",
-      "target": ".codex/skills/superspec-review/SKILL.md"
-    },
-    {
-      "kind": "skill",
-      "source": "templates/workflow/skills/superspec-archive/SKILL.md",
-      "target": ".codex/skills/superspec-archive/SKILL.md"
-    },
-    {
-      "kind": "hook",
-      "source": "templates/hooks/codex-hooks.json",
-      "target": ".codex/hooks.json"
-    },
-    {
-      "kind": "prompt",
-      "source": "templates/workflow/prompts/architect.md",
-      "target": ".codex/prompts/architect.md"
-    },
-    {
-      "kind": "prompt",
-      "source": "templates/workflow/prompts/code-reviewer.md",
-      "target": ".codex/prompts/code-reviewer.md"
-    },
-    {
-      "kind": "prompt",
-      "source": "templates/workflow/prompts/critic.md",
-      "target": ".codex/prompts/critic.md"
-    },
-    {
-      "kind": "prompt",
-      "source": "templates/workflow/prompts/executor.md",
-      "target": ".codex/prompts/executor.md"
-    },
-    {
-      "kind": "prompt",
-      "source": "templates/workflow/prompts/explore.md",
-      "target": ".codex/prompts/explore.md"
-    },
-    {
-      "kind": "prompt",
-      "source": "templates/workflow/prompts/test-runner.md",
-      "target": ".codex/prompts/test-runner.md"
-    },
-    {
-      "kind": "prompt",
-      "source": "templates/workflow/prompts/test-engineer.md",
-      "target": ".codex/prompts/test-engineer.md"
-    },
-    {
-      "kind": "prompt",
-      "source": "templates/workflow/prompts/verifier.md",
-      "target": ".codex/prompts/verifier.md"
-    },
-    {
-      "kind": "agent",
-      "source": "adapters/codex/agents/architect.toml",
-      "target": ".codex/agents/architect.toml"
-    },
-    {
-      "kind": "agent",
-      "source": "adapters/codex/agents/code-reviewer.toml",
-      "target": ".codex/agents/code-reviewer.toml"
-    },
-    {
-      "kind": "agent",
-      "source": "adapters/codex/agents/critic.toml",
-      "target": ".codex/agents/critic.toml"
-    },
-    {
-      "kind": "agent",
-      "source": "adapters/codex/agents/explore.toml",
-      "target": ".codex/agents/explore.toml"
-    },
-    {
-      "kind": "agent",
-      "source": "adapters/codex/agents/executor.toml",
-      "target": ".codex/agents/executor.toml"
-    },
-    {
-      "kind": "agent",
-      "source": "adapters/codex/agents/test-runner.toml",
-      "target": ".codex/agents/test-runner.toml"
-    },
-    {
-      "kind": "agent",
-      "source": "adapters/codex/agents/test-engineer.toml",
-      "target": ".codex/agents/test-engineer.toml"
-    },
-    {
-      "kind": "agent",
-      "source": "adapters/codex/agents/verifier.toml",
-      "target": ".codex/agents/verifier.toml"
-    }
-  ]
-}

package/bin/launch.js

@@ -1,37 +0,0 @@
-export const REQUIRED_NODE_VERSION = "20.19.0";
-const REQUIRED_NODE_PARTS = REQUIRED_NODE_VERSION.split(".").map((part) => Number.parseInt(part, 10));
-
-function compareNodeVersion(version) {
-  const parts = String(version).split(".").map((part) => Number.parseInt(part, 10));
-  for (let index = 0; index < REQUIRED_NODE_PARTS.length; index += 1) {
-    const actual = parts[index] ?? 0;
-    const required = REQUIRED_NODE_PARTS[index] ?? 0;
-    if (!Number.isFinite(actual)) return -1;
-    if (actual > required) return 1;
-    if (actual < required) return -1;
-  }
-  return 0;
-}
-
-export function nodeVersionError(version = process.versions.node) {
-  if (compareNodeVersion(version) >= 0) return null;
-  return `SuperSpec requires Node.js >=${REQUIRED_NODE_VERSION}. Current Node.js is ${version}. Upgrade Node.js, then rerun the command.`;
-}
-
-export function assertSupportedNode(version = process.versions.node) {
-  const message = nodeVersionError(version);
-  if (message === null) return;
-  console.error(message);
-  process.exit(1);
-}
-
-export async function runEntry(entryPath, exportName, argv = process.argv.slice(2)) {
-  assertSupportedNode();
-  const mod = await import(new URL(entryPath, import.meta.url).href);
-  const main = mod[exportName];
-  if (typeof main !== "function") {
-    console.error(`SuperSpec internal error: ${entryPath} does not export ${exportName}`);
-    process.exit(2);
-  }
-  process.exitCode = await main(argv);
-}

package/bin/superspec-check.js

@@ -1,4 +0,0 @@
-#!/usr/bin/env node
-import { runEntry } from "./launch.js";
-
-await runEntry("../dist/superspec_guard.js", "main");

package/bin/superspec-hook.js

@@ -1,4 +0,0 @@
-#!/usr/bin/env node
-import { runEntry } from "./launch.js";
-
-await runEntry("../dist/superspec_hook.js", "main_hook");

package/bin/superspec-init.js

@@ -1,4 +0,0 @@
-#!/usr/bin/env node
-import { runEntry } from "./launch.js";
-
-await runEntry("../dist/superspec_init.js", "main_init_async");

package/bin/superspec.js

@@ -1,4 +0,0 @@
-#!/usr/bin/env node
-import { runEntry } from "./launch.js";
-
-await runEntry("../dist/superspec.js", "main_superspec");