harnessed 3.4.3 → 3.5.0

package/workflows/task/auto/SKILL.md

@@ -22,7 +22,7 @@ trigger_phrases:
 ## Overview
 
 4-stage cadence Stage ③ master orchestrator delegating to 4 sequential sub-workflows
-per subtask (per ~/.claude/CLAUDE.md "Execute 阶段" 节 + karpathy 心法 always-on):
+per subtask (bundled Execute-stage cadence + karpathy 心法 always-on):
 
 | order | sub | gate ref | mode | when fires |
 | ----- | --- | -------- | ---- | ---------- |
@@ -39,7 +39,7 @@ mode delegate carries explicit `order`。Each subtask 入口走一次此 master
 
 ralph-loop 是正交 wrapper, 套在 deliver sub 的 01-deliver phase 外层保 completion-promise
 verbatim "COMPLETE" (R20.10)。任何执行单元 (subagent / team / 主 session) 都可外层套 ralph-loop
-保 completion-promise (sister ~/.claude/CLAUDE.md 子任务并行执行机制 "正交 wrapper")。
+保 completion-promise (bundled subagent vs Agent Teams routing — orthogonal wrapper rule).
 
 ## Capability refs
 
@@ -57,20 +57,24 @@ Sister `workflows/capabilities.yaml`:
 
 - Slash command: `/task <text>` (bare per ADR 0030 namespace policy D-02 LOCK after `harnessed setup`)
 
-<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-**Preferred path** (master orchestrator): dispatch to the per-sub-workflow slash commands in the order this stage prescribes. Each sub command lives at `~/.claude/commands/<sub-name>.md` with its own dual-path fallback.
+Use the Bash tool to run:
 
-**Fallback path** (when no slash command from the sub-list resolves): run each missing sub-workflow inline using its own role prompt from `~/.claude/skills/<sub-name>/SKILL.md`. Do NOT skip stages silently — each sub either runs or is logged as "skipped: <reason>".
+```bash
+echo "$ARGUMENTS" | harnessed run task --task-stdin
+```
 
-(Sister `~/.claude/commands/task.md` is also generated by `harnessed setup` so `/task` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed task --apply` CLI claims are removed; that subcommand was never implemented.)
+If `$ARGUMENTS` is empty, run `harnessed run task` (no stdin pipe).
+
+After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
+
+<!-- harnessed-generated:v3.4.4 -->
 
 ## References
 
 - D-01 master orchestrator delegation pattern
 - D-02 bare slash cmd convention (ADR 0030 namespace policy LOCK)
 - D-10 ralph-loop orthogonal wrapper
-- ~/.claude/CLAUDE.md "Execute 阶段" 节 verbatim (karpathy 心法 + mattpocock 招式 + ralph-loop COMPLETE)
 - workflows/judgments/{subtask-gate,tdd-gate}.yaml — brainstorming + tdd-strongly-suggested triggers
 - workflows/task/{clarify,code,test,deliver}/workflow.yaml — 4 sub-workflow Phase 3.4 SHIPPED

package/workflows/task/auto/workflow.yaml

@@ -1,13 +1,12 @@
 # workflows/task/auto/workflow.yaml — Phase v3.0-3.5 W1 T3.5.W1.3
 #
 # Stage ③ Task master orchestrator — 串行 invoke 4 sub per subtask (clarify → code → test → deliver)
-# per ~/.claude/CLAUDE.md "Execute 阶段" 节 + RESEARCH-workflows § Area 2。
+# per bundled Execute-stage cadence + RESEARCH-workflows § Area 2。
 # ralph-loop COMPLETE wrapper 在 deliver phase 内 (D-10 orthogonal wrapper not master concern)。
 # tdd-gate conditional fire on test sub (sister task/test/workflow.yaml gate
 # judgments.tdd-gate.tdd-strongly-suggested.fires)。
 #
 # Sister refs:
-#   - ~/.claude/CLAUDE.md "Execute 阶段 (GSD orchestration + superpowers 子任务执行)" 节
 #   - workflows/judgments/subtask-gate.yaml — brainstorming trigger (clarify)
 #   - workflows/judgments/tdd-gate.yaml — tdd-strongly-suggested trigger (test conditional)
 #   - workflows/task/{clarify,code,test,deliver}/workflow.yaml — 4 sub-workflow Phase 3.4 SHIPPED
@@ -27,7 +26,7 @@ description: |
   Stage ③ Task master orchestrator — 串行 invoke 4 sub per subtask (clarify → code → test → deliver);
   ralph-loop COMPLETE wrapper 在 deliver phase 内 (D-10 orthogonal wrapper not master concern)。
   tdd-gate conditional fire on test sub (核心业务 / 算法 / 数据处理 / 回归 risk high / reliability
-  required, 6 fires_when OR-chain per ~/.claude/CLAUDE.md)。execute-task 每个 subtask 入口走一次。
+  required, 6 fires_when OR-chain; bundled tdd-gate judgment)。execute-task 每个 subtask 入口走一次。
 
 disciplines_applied: [karpathy, output-style, language, operational, priority, protocols]
 tools_available:

package/workflows/task/clarify/SKILL.md

@@ -54,37 +54,19 @@ sister CLAUDE.md "Discuss / Research 阶段" mattpocock 招式按需召唤 patte
 unconditional fire (D-05 invokes_tools 与 OnClause 并存, 但作用面不同 — invokes_tools
 phase-level conditional tool fire NOT 决定 phase 是否走)。
 
-<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-**Preferred path** (when the upstream specialist is installed): use the SlashCommand tool to run `{{ capabilities.superpowers-brainstorming.cmd }}` — the upstream specialist takes over.
-
-**Fallback path** (when the upstream isn't installed or returns no result): use the Task tool to spawn a general-purpose subagent with this prompt:
-
-> You are a **Subtask spec clarifier**.
->
-> **Mission**: Surface ambiguity in a single subtask spec by asking ONE focused question at a time. Fires when ≥2 approaches / core algorithm / API contract / high error-cost. Skip if subtask is CRUD or already obvious.
->
-> **Default-suspect mode**: assume the change is broken / risky / incomplete until proven otherwise. Cite `file:line` for every finding; do not generalize.
->
-> **Review checklist**:
-> 1. Read the subtask description; restate it in your own words to confirm
->
-> 2. List every assumption you would make; flag the ones the user must confirm
->
-> 3. Ask ONE question at a time, lowest-cost-to-answer first
->
-> 4. Stop asking when you have enough to write 80% of the code without guessing
->
-> 5. Record the resolved spec at the top of the subtask file before implementing
->
-> 6. If `phase.spec_ambiguous == true AND phase.no_docs == true`, request grill-me
->
-> **Output format**: structured report with severity-classified findings (blocking-question / nice-to-know / resolved). One finding per line: `[severity] file:line — problem (one sentence); fix: suggested change`. If no findings, say so explicitly. No preamble, no end-of-report summary.
-
-(Role prompt is self-contained — works even when the upstream `superpowers-brainstorming` user-skill / plugin isn't installed.)
-
-(Sister `~/.claude/commands/task-clarify.md` is also generated by `harnessed setup` so `/task-clarify` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed task-clarify --apply` CLI claims are removed; that subcommand was never implemented.)
+Use the Bash tool to run:
+
+```bash
+echo "$ARGUMENTS" | harnessed run task-clarify --task-stdin
+```
+
+If `$ARGUMENTS` is empty, run `harnessed run task-clarify` (no stdin pipe).
+
+After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
+
+<!-- harnessed-generated:v3.4.4 -->
 
 ## References
 
@@ -92,7 +74,6 @@ phase-level conditional tool fire NOT 决定 phase 是否走)。
 - D-05 — phase-level `invokes_tools` conditional tool fire
 - D-04 — `gate` 4-level ref pre-resolved by `judgmentResolver`
 - D-02 — SKILL.md `name:` bare slash cmd (`task-clarify` NOT `task/clarify`) per ADR 0030
-- ~/.claude/CLAUDE.md "Execute 阶段" 节 brainstorming 子任务层判据
 - `workflows/judgments/subtask-gate.yaml` triggers.brainstorming
 - `workflows/capabilities.yaml` — superpowers-brainstorming + grill-with-docs entries
 - `workflows/defaults.yaml` — ralph_max_iterations.task-clarify.* values (T3.4.W2.2 followup)

package/workflows/task/code/SKILL.md

@@ -56,43 +56,23 @@ Phase 01-code 按 phase fact context 条件性 fire 3 mattpocock 招式:
 
 02-progress invokes the **Claude Code plugin** slash command `/plan` to update
 `progress.md` in `.planning/<phase-id>/` — 跟踪 subtask 完成 / blocked / next step
-per CLAUDE.md "跨 session 恢复" 模式 + R20.6 Manus-style 持久化。Plugin path
-verified at `~/.claude/plugins/cache/planning-with-files/planning-with-files/2.34.0/`
-(2026-05-20).
+per bundled "跨 session 恢复" 模式 + R20.6 Manus-style 持久化. Requires the
+`planning-with-files` Claude Code plugin (install via Claude Code plugin
+marketplace).
 
-<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-**Preferred path** (when the upstream specialist is installed): use the SlashCommand tool to run `{{ capabilities.planning-with-files.cmd }}` — the upstream specialist takes over.
-
-**Fallback path** (when the upstream isn't installed or returns no result): use the Task tool to spawn a general-purpose subagent with this prompt:
-
-> You are a **Karpathy-discipline implementer**.
->
-> **Mission**: Implement a single subtask under karpathy 4 心法 (Think Before Coding / Simplicity First / Surgical Changes / Goal-Driven Execution) with ≤200 LOC per file. Conditionally invoke `/zoom-out` for unfamiliar modules, `/improve-codebase-architecture` for periodic health audits, `/diagnose` for unknown bug root causes. Update `progress.md` via planning-with-files `/plan` when done.
->
-> **Default-suspect mode**: assume the change is broken / risky / incomplete until proven otherwise. Cite `file:line` for every finding; do not generalize.
->
-> **Review checklist**:
-> 1. Before any edit: read the file you intend to change end-to-end
->
-> 2. Smallest change that satisfies the acceptance criteria — no scope creep
->
-> 3. ≤200 LOC per file (split modules if growing past it)
->
-> 4. Trust internal code: don't re-validate already-checked inputs at every layer
->
-> 5. No speculative abstractions (no 'just in case' generics)
->
-> 6. Edit with surgical precision: full path, exact selectors, no broad rewrites
->
-> 7. Update progress.md before declaring done (planning-with-files `/plan`)
->
-> **Output format**: structured report with severity-classified findings (needs-fix / done / blocked). One finding per line: `[severity] file:line — problem (one sentence); fix: suggested change`. If no findings, say so explicitly. No preamble, no end-of-report summary.
-
-(Role prompt is self-contained — works even when the upstream `planning-with-files` user-skill / plugin isn't installed.)
-
-(Sister `~/.claude/commands/task-code.md` is also generated by `harnessed setup` so `/task-code` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed task-code --apply` CLI claims are removed; that subcommand was never implemented.)
+Use the Bash tool to run:
+
+```bash
+echo "$ARGUMENTS" | harnessed run task-code --task-stdin
+```
+
+If `$ARGUMENTS` is empty, run `harnessed run task-code` (no stdin pipe).
+
+After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
+
+<!-- harnessed-generated:v3.4.4 -->
 
 ## References
 
@@ -100,7 +80,6 @@ verified at `~/.claude/plugins/cache/planning-with-files/planning-with-files/2.3
 - D-05 — phase-level `invokes_tools` conditional tool fire
 - D-15 + Q-AUDIT-5a — planning-with-files = Claude Code plugin slash cmd `/plan`
 - D-02 — SKILL.md `name:` bare slash cmd (`task-code` NOT `task/code`) per ADR 0030
-- ~/.claude/CLAUDE.md "Execute 阶段" 节 karpathy 心法 + mattpocock 招式按需召唤
 - `workflows/disciplines/karpathy.yaml` — 4 心法 + ≤200L hard limit 等 rules (L0 substrate)
 - `workflows/capabilities.yaml` — zoom-out / improve-codebase-architecture / diagnose / planning-with-files entries
 - `workflows/defaults.yaml` — ralph_max_iterations.task-code.* values (T3.4.W2.2 followup)

package/workflows/task/code/workflow.yaml

@@ -8,7 +8,6 @@
 #
 # Sister refs:
 #   - .planning/phase-v3.0-3.2/RESEARCH-workflows.md § Area 2 task/code sub yaml example
-#   - ~/.claude/CLAUDE.md "Execute 阶段" 节 karpathy 4 心法 + mattpocock 招式按需召唤
 #   - workflows/capabilities.yaml zoom-out + improve-codebase-architecture + diagnose +
 #     planning-with-files (Phase 2.3 W0.1 SHIPPED)
 #   - workflows/disciplines/karpathy.yaml (Phase 3.3 W0.4 SHIPPED, L0 substrate always-on)

package/workflows/task/deliver/SKILL.md

@@ -41,7 +41,7 @@ spawns each phase as a sub-agent via `@anthropic-ai/claude-agent-sdk` 0.3.142+.
 ralph-loop SDK wrapper 保 completion-promise verbatim string `"COMPLETE"` — sub-task
 被认为完成的判据是子任务输出包含 verbatim "COMPLETE" string (NOT 启发式 / NOT
 LLM-as-judge). Sister capabilities.yaml `ralph-loop` entry impl `bundled-skill` +
-`sdk_ref: src/routing/lib/ralphLoop.ts` (Phase 2.2 v0.2.0 ship)。
+`sdk_ref: src/workflow/lib/ralphLoop.ts` (Phase 2.2 v0.2.0 ship)。
 
 ### Parallelism — ralph-loop 正交 wrapper
 
@@ -79,55 +79,32 @@ ref / ✅ R20.10 explicit max_iterations_exceeded handler。
 
 02-progress-mark invokes Claude Code plugin slash cmd `/plan` to mark subtask complete
 in `progress.md` — sister Phase 01-code progress update pattern, last call in Stage
-③ task chain。Plugin path `~/.claude/plugins/cache/planning-with-files/
-planning-with-files/2.34.0/` verified (2026-05-20)。
+③ task chain. Requires `planning-with-files` Claude Code plugin (install via
+Claude Code plugin marketplace).
 
-<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-**Preferred path** (when the upstream specialist is installed): use the SlashCommand tool to run `{{ capabilities.ralph-loop.cmd }}` — the upstream specialist takes over.
-
-**Fallback path** (when the upstream isn't installed or returns no result): use the Task tool to spawn a general-purpose subagent with this prompt:
-
-> You are a **Completion-promise enforcer (ralph-loop COMPLETE)**.
->
-> **Mission**: Wrap the subtask in ralph-loop with `completion_promise: "COMPLETE"` and `max_iterations: <N>`. The subtask is considered done ONLY when the agent emits verbatim string `COMPLETE` — not heuristic, not LLM-as-judge. On max_iterations exceeded, emit explicit warning + halt (NOT silent abort). Then mark progress.md complete.
->
-> **Default-suspect mode**: assume the change is broken / risky / incomplete until proven otherwise. Cite `file:line` for every finding; do not generalize.
->
-> **Review checklist**:
-> 1. Confirm subtask acceptance criteria are explicit and verifiable BEFORE looping
->
-> 2. Set `max_iterations` based on subtask size; default 20
->
-> 3. On loop entry, give the agent the full spec + acceptance criteria + completion promise
->
-> 4. If agent emits 'COMPLETE' verbatim, mark progress.md done via `/plan`
->
-> 5. If max_iterations exceeded, emit warning + halt; do NOT silent-continue
->
-> 6. If teammate communication needed / context overflow → escalate to Agent Teams
->
-> 7. Cleanup: SendMessage shutdown_request + TeamDelete (防呆清单 mandatory)
->
-> **Output format**: structured report with severity-classified findings (complete / max-iter-exceeded / escalated-to-teams). One finding per line: `[severity] file:line — problem (one sentence); fix: suggested change`. If no findings, say so explicitly. No preamble, no end-of-report summary.
-
-(Role prompt is self-contained — works even when the upstream `ralph-loop` user-skill / plugin isn't installed.)
-
-(Sister `~/.claude/commands/task-deliver.md` is also generated by `harnessed setup` so `/task-deliver` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed task-deliver --apply` CLI claims are removed; that subcommand was never implemented.)
+Use the Bash tool to run:
+
+```bash
+echo "$ARGUMENTS" | harnessed run task-deliver --task-stdin
+```
+
+If `$ARGUMENTS` is empty, run `harnessed run task-deliver` (no stdin pipe).
+
+After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
+
+<!-- harnessed-generated:v3.4.4 -->
 
 ## References
 
 - D-09 — L0 Discipline Substrate always-on (6 disciplines)
 - D-10 — ralph-loop 真接 SDK wrapper (NOT mock reference; v0.2.0 ship)
-- D-11 — Agent Teams 升级 5 触发 OR-chain per ~/.claude/rules/agent-teams.md
+- D-11 — Agent Teams 升级 5 触发 OR-chain per bundled parallelism-gate rules
 - R20.10 — ralph-loop max_iterations_exceeded explicit emit_warning_and_halt
   (acceptance c "NOT silent abort"); ralph-loop 正交 wrapper wraps 3 mode
 - D-02 — SKILL.md `name:` bare slash cmd (`task-deliver` NOT `task/deliver`) per ADR 0030
 - ADR 0011 — SDK + ralph-loop integration v0.2.0 baseline
-- ~/.claude/CLAUDE.md "Execute 阶段" 节 ralph-loop 至 verbatim COMPLETE +
-  `子任务并行执行机制` Agent Teams 升级路由
-- ~/.claude/rules/agent-teams.md — Pattern A/B/C + 防呆清单 cleanup mandatory
 - `workflows/judgments/parallelism-gate.yaml` triggers.ralph-loop-wrapper +
   agent-teams-upgrade + subagent-default + main-session-fallback
 - `workflows/capabilities.yaml` — ralph-loop + agent-teams-{create,send-message,shutdown}

package/workflows/task/deliver/workflow.yaml

@@ -3,8 +3,8 @@
 #
 # Stage ③.d 子任务交付 — ralph-loop COMPLETE wrapper 保 completion-promise verbatim
 # "COMPLETE" (R20.10 + D-10 v0.2.0 ralph-loop SDK pattern)。Parallelism: subagent
-# default fan-out OR Agent Teams 升级 (5 触发 OR-chain per ~/.claude/rules/agent-teams.md
-# + D-11)。02-progress-mark: planning-with-files /plan 更新 progress.md 标记完成。
+# default fan-out OR Agent Teams 升级 (5 触发 OR-chain per bundled Agent Teams
+# routing + D-11)。02-progress-mark: planning-with-files /plan 更新 progress.md 标记完成。
 #
 # Brief enforcement (W0.9 spec MUST include):
 #   - ralph-loop completion_promise = COMPLETE (R20.10 verbatim acceptance string)
@@ -20,8 +20,9 @@
 #   - workflows/judgments/parallelism-gate.yaml triggers.ralph-loop-wrapper (Phase 2.3 W0.2 SHIPPED)
 #   - workflows/capabilities.yaml ralph-loop + agent-teams-{create,send-message,shutdown} +
 #     planning-with-files (Phase 2.3 W0.1 + 3.3 W0.2 SHIPPED)
-#   - ~/.claude/rules/agent-teams.md 5 升级触发 OR-chain (teammate_send_message_needed /
-#     subagent_context_overflow / shared_task_list / opposing_hypothesis_debate / fullstack_three_way)
+#   - workflows/judgments/parallelism-gate.yaml — 5 升级触发 OR-chain
+#     (teammate_send_message_needed / subagent_context_overflow / shared_task_list /
+#     opposing_hypothesis_debate / fullstack_three_way)
 #   - src/workflow/schema/workflow.ts WorkflowSchemaV3 FallbackMaxIterationsExceeded
 #     Type.Literal('emit_warning_and_halt') (Phase 3.3 W0.5 SHIPPED)
 #
@@ -33,11 +34,11 @@ workflow: task-deliver
 description: |
   Stage ③.d 子任务交付 — ralph-loop COMPLETE wrapper 保 completion-promise verbatim
   "COMPLETE" (R20.10 + D-10 v0.2.0 ralph-loop SDK pattern)。Parallelism subagent
-  default fan-out OR Agent Teams 升级 (5 触发 OR-chain per ~/.claude/rules/agent-teams.md
+  default fan-out OR Agent Teams 升级 (5 触发 OR-chain per bundled Agent Teams routing
   + D-11)。R20.10 explicit max_iterations_exceeded handler: emit_warning_and_halt
   exit_code 1 (NOT silent abort, sister Phase 2.4 W1.2 fallbackHandlers.ts engine.ts
   wire pattern)。Cleanup mandatory when Agent Teams escalation path fires (shutdown_request
-  + TeamDelete per agent-teams.md 防呆清单)。
+  + TeamDelete per bundled Agent Teams cleanup discipline)。
 
 disciplines_applied: [karpathy, output-style, language, operational, priority, protocols]
 tools_available:

package/workflows/task/test/SKILL.md

@@ -63,39 +63,19 @@ Phase 01-test 条件性 fire `diagnose` (capabilities.yaml L55-64 mattpocock-ski
 test fail 时进入 diagnose loop (reproduce → minimise → hypothesise → instrument →
 fix → regression-test), 测试通过则 skip diagnose entirely。
 
-<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-**Preferred path** (when the upstream specialist is installed): use the SlashCommand tool to run `{{ capabilities.tdd.cmd }}` — the upstream specialist takes over.
-
-**Fallback path** (when the upstream isn't installed or returns no result): use the Task tool to spawn a general-purpose subagent with this prompt:
-
-> You are a **TDD enforcer (red-green-refactor)**.
->
-> **Mission**: Drive red-green-refactor for core business logic / algorithms / data processing / regression-risk / reliability-required subtasks. Skip pure CRUD / UI polish / docs-only. On test failure, hand off to `/diagnose` for systematic root-cause.
->
-> **Default-suspect mode**: assume the change is broken / risky / incomplete until proven otherwise. Cite `file:line` for every finding; do not generalize.
->
-> **Review checklist**:
-> 1. Red: write ONE failing test for the smallest behavior increment; run, watch it fail
->
-> 2. Green: write the minimum code that makes it pass — nothing more
->
-> 3. Refactor: clean up duplication / clarify names — keep tests green
->
-> 4. Loop. Each cycle ≤10 min; if longer, the increment is too big — split
->
-> 5. Negative cases matter: at least 1 test per error / edge / boundary
->
-> 6. Test name = expected behavior, not 'test1', not 'should work'
->
-> 7. On unexpected failure: stop adding tests; route to `/diagnose` for root cause
->
-> **Output format**: structured report with severity-classified findings (red / green / refactored / blocked). One finding per line: `[severity] file:line — problem (one sentence); fix: suggested change`. If no findings, say so explicitly. No preamble, no end-of-report summary.
-
-(Role prompt is self-contained — works even when the upstream `tdd` user-skill / plugin isn't installed.)
-
-(Sister `~/.claude/commands/task-test.md` is also generated by `harnessed setup` so `/task-test` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed task-test --apply` CLI claims are removed; that subcommand was never implemented.)
+Use the Bash tool to run:
+
+```bash
+echo "$ARGUMENTS" | harnessed run task-test --task-stdin
+```
+
+If `$ARGUMENTS` is empty, run `harnessed run task-test` (no stdin pipe).
+
+After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
+
+<!-- harnessed-generated:v3.4.4 -->
 
 ## References
 
@@ -104,7 +84,6 @@ fix → regression-test), 测试通过则 skip diagnose entirely。
 - D-05 — phase-level `invokes_tools` conditional tool fire
 - D-13 — tdd capability 2 impl 候选 alias (superpowers 主 + mattpocock /tdd 备)
 - D-02 — SKILL.md `name:` bare slash cmd (`task-test` NOT `task/test`) per ADR 0030
-- ~/.claude/CLAUDE.md "Execute 阶段" TDD 强烈建议开启 6 触发 + 3 跳过
 - `workflows/judgments/tdd-gate.yaml` triggers.tdd-strongly-suggested
 - `workflows/capabilities.yaml` — tdd (superpowers + mattpocock alias) + diagnose entries
 - `workflows/defaults.yaml` — ralph_max_iterations.task-test.* values (T3.4.W2.2 followup)

package/workflows/task/test/workflow.yaml

@@ -9,7 +9,6 @@
 #   - .planning/phase-v3.0-3.2/RESEARCH-workflows.md § Area 2 task/test sub yaml example
 #   - workflows/judgments/tdd-gate.yaml triggers.tdd-strongly-suggested (Phase 2.3 W0.2 SHIPPED)
 #   - workflows/capabilities.yaml tdd + diagnose (Phase 2.3 W0.1 SHIPPED + D-13 alias)
-#   - ~/.claude/CLAUDE.md "Execute 阶段" 节 TDD 强烈建议开启 6 触发 + 3 跳过
 #   - src/workflow/schema/workflow.ts WorkflowSchemaV3 (Phase 3.3 W0.5 SHIPPED)
 #
 # Brief enforcement:
@@ -24,7 +23,7 @@ workflow: task-test
 description: |
   Stage ③.c 子任务测试 — superpowers TDD red-green-refactor 强制 (核心业务 / 算法 /
   数据处理 / 回归 risk high / reliability required, 6 fires_when OR-chain per
-  ~/.claude/CLAUDE.md), 其他可选 mattpocock /tdd alias。Conditional diagnose
+  bundled tdd-gate judgment), 其他可选 mattpocock /tdd alias。Conditional diagnose
   invoke when test_fail (D-05 invokes_tools phase-level fire). Gate:
   judgments.tdd-gate.tdd-strongly-suggested.fires (4-level ref, pre-resolved per D-04)。
 

package/workflows/verify/auto/SKILL.md

@@ -1,8 +1,8 @@
 ---
 name: verify
 description: |
-  Stage ④ Verify master orchestrator — 7 sub conditional per ~/.claude/CLAUDE.md "Verify 阶段"
-  verbatim: progress 必跑 → code-review 并行 → paranoid 关键模块强制 → qa/security/design 可选
+  Stage ④ Verify master orchestrator — 7 sub conditional per bundled Verify-stage cadence:
+  progress 必跑 → code-review 并行 → paranoid 关键模块强制 → qa/security/design 可选
   并行 conditional → simplify 末尾 → multispec 关键发布 Pattern C 4-specialist Agent Team。
   schema_version: harnessed.workflow.v3 with delegates_to (7 sub: progress serial order 1 +
   5 parallel conditional + simplify serial order 99) + disciplines_applied (6 default) +
@@ -21,8 +21,8 @@ trigger_phrases:
 ## Overview
 
 4-stage cadence Stage ④ master orchestrator delegating to 7 sub-workflows
-(per ~/.claude/CLAUDE.md "Verify 阶段" 节 verbatim — 9-phase composition compressed
-into 7 sub delegation via stage-routing.yaml):
+(bundled Verify-stage cadence — 9-phase composition compressed into 7 sub delegation
+via stage-routing.yaml):
 
 | order/mode | sub | gate ref | mode | when fires |
 | ---------- | --- | -------- | ---- | ---------- |
@@ -64,22 +64,26 @@ Sister `workflows/capabilities.yaml`:
 
 - Slash command: `/verify` (bare per ADR 0030 namespace policy D-02 LOCK after `harnessed setup`)
 
-<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-**Preferred path** (master orchestrator): dispatch to the per-sub-workflow slash commands in the order this stage prescribes. Each sub command lives at `~/.claude/commands/<sub-name>.md` with its own dual-path fallback.
+Use the Bash tool to run:
 
-**Fallback path** (when no slash command from the sub-list resolves): run each missing sub-workflow inline using its own role prompt from `~/.claude/skills/<sub-name>/SKILL.md`. Do NOT skip stages silently — each sub either runs or is logged as "skipped: <reason>".
+```bash
+echo "$ARGUMENTS" | harnessed run verify --task-stdin
+```
 
-(Sister `~/.claude/commands/verify.md` is also generated by `harnessed setup` so `/verify` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed verify --apply` CLI claims are removed; that subcommand was never implemented.)
+If `$ARGUMENTS` is empty, run `harnessed run verify` (no stdin pipe).
+
+After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
+
+<!-- harnessed-generated:v3.4.4 -->
 
 ## References
 
 - D-01 master orchestrator delegation pattern
 - D-02 bare slash cmd convention (ADR 0030 namespace policy LOCK)
 - D-12 gstack 治理关卡 ref (paranoid / qa / security / design subs)
-- ~/.claude/CLAUDE.md "Verify 阶段" 节 verbatim
-- ~/.claude/rules/agent-teams.md Pattern C 多维度审查 (multispec sub 4-specialist 互相质询)
+- workflows/judgments/parallelism-gate.yaml — Pattern C 多维度审查 (multispec sub 4-specialist 互相质询)
 - workflows/judgments/stage-routing.yaml — verify-* 6 triggers (7 sub delegation)
 - workflows/verify/{progress,code-review,paranoid,qa,security,design,simplify,multispec}/workflow.yaml
   — 8 sub-workflow Phase 3.4 SHIPPED

package/workflows/verify/auto/workflow.yaml

@@ -1,7 +1,7 @@
 # workflows/verify/auto/workflow.yaml — Phase v3.0-3.5 W1 T3.5.W1.4
 #
-# Stage ④ Verify master orchestrator — 7 sub conditional per ~/.claude/CLAUDE.md "Verify 阶段"
-# verbatim + RESEARCH-workflows § Area 2 verify master:
+# Stage ④ Verify master orchestrator — 7 sub conditional per bundled Verify-stage cadence
+# + RESEARCH-workflows § Area 2 verify master:
 #   progress 必跑 (serial order 1) →
 #   code-review 并行 (parallel) +
 #   paranoid 关键模块 (parallel + gate is-critical-module) +
@@ -10,7 +10,6 @@
 #   multispec 关键发布 Pattern C (parallel + gate critical-release-upgrade)
 #
 # Sister refs:
-#   - ~/.claude/CLAUDE.md "Verify 阶段 (GSD + gstack)" 节 verbatim
 #   - workflows/judgments/stage-routing.yaml — verify-* triggers (7 sub delegation)
 #   - workflows/judgments/parallelism-gate.yaml — agent-teams-upgrade trigger (multispec gate)
 #   - workflows/verify/{progress,code-review,paranoid,qa,security,design,simplify,multispec}/workflow.yaml
@@ -30,8 +29,8 @@
 schema_version: harnessed.workflow.v3
 workflow: verify
 description: |
-  Stage ④ Verify master orchestrator — 7 sub conditional per ~/.claude/CLAUDE.md "Verify 阶段"
-  verbatim: progress 必跑串行 → code-review 并行 → paranoid 关键模块强制 → qa/security/design
+  Stage ④ Verify master orchestrator — 7 sub conditional per bundled Verify-stage cadence:
+  progress 必跑串行 → code-review 并行 → paranoid 关键模块强制 → qa/security/design
   可选 conditional 并行 → simplify 末尾串行 → multispec 关键发布 Pattern C 4-specialist Agent Team。
   serial mode 必带 explicit order (K9 invariant); progress order 1 起点, simplify order 99 末尾。
 

package/workflows/verify/code-review/SKILL.md

@@ -2,7 +2,7 @@
 name: verify-code-review
 description: |
   Stage ④.b verify sub-workflow — code-review 多 agent 并行 fan-out 高置信度 finding
-  (subagent default per ~/.claude/CLAUDE.md 子任务并行机制 — Task / Agent 工具 spawn
+  (subagent default per bundled parallelism gate — Task / Agent 工具 spawn
   多 subagent fan-out, context 隔离, token 敏感)。
   schema_version: harnessed.workflow.v3 with disciplines_applied (6 default) + tools_available
   (code-review) + 1 phase (parallelism ref judgments.parallelism-gate.subagent-default.fires)。
@@ -29,8 +29,8 @@ onto harnessed runtime (Phase v3.0-3.4 W0.11 — D-04 Stage ④ Verify 7 sub +
 | 1 | `01-code-review` | mattpocock-skills | sonnet | `{{ capabilities.code-review.cmd }}` | `judgments.parallelism-gate.subagent-default.fires` |
 
 Per-phase config loads from `workflows/verify/code-review/workflow.yaml`; engine spawns
-multiple subagent in parallel fan-out (sister `~/.claude/CLAUDE.md` 子任务并行机制 默认 —
-Task / Agent 工具 spawn 多任务并发, context 隔离, summary 折叠回主 context)。
+multiple subagent in parallel fan-out (bundled subagent-default rule — Task / Agent
+工具 spawn 多任务并发, context 隔离, summary 折叠回主 context)。
 
 ## Capability refs
 
@@ -48,47 +48,23 @@ Sister `workflows/judgments/parallelism-gate.yaml`:
 总 fire 当 `phase.stage == 'verify'` 后必跑串行 (verify-progress) 之后并行 fan-out。无 skip
 条件 — code-review 多 agent 是 verify-work 第 3 phase 默认 fan-out (sister CLAUDE.md verbatim)。
 
-<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-**Preferred path** (when the upstream specialist is installed): use the SlashCommand tool to run `{{ capabilities.code-review.cmd }}` — the upstream specialist takes over.
-
-**Fallback path** (when the upstream isn't installed or returns no result): use the Task tool to spawn a general-purpose subagent with this prompt:
-
-> You are a **Code Reviewer (multi-agent fan-out)**.
->
-> **Mission**: Spawn parallel sonnet agents that each review the diff from a different angle (CLAUDE.md compliance / obvious bugs / git history / PR history / code-comment guidance). Filter findings by confidence ≥80. Adapted from claude-plugins-official `code-review` plugin pattern.
->
-> **Default-suspect mode**: assume the change is broken / risky / incomplete until proven otherwise. Cite `file:line` for every finding; do not generalize.
->
-> **Review checklist**:
-> 1. Read the diff against the base branch — full diff, not just summaries
->
-> 2. Audit against CLAUDE.md (root + any directory-level CLAUDE.md)
->
-> 3. Shallow scan for obvious bugs in changed lines (avoid context expansion)
->
-> 4. Git blame on modified regions — bugs visible only in historical context
->
-> 5. Previous PRs touching same files — recurring patterns / past comments
->
-> 6. Inline code comments / docstrings — does the change violate stated invariants?
->
-> 7. Score each finding 0-100; drop <80; cite file:line for kept findings
->
-> 8. Avoid: pre-existing issues, linter-catchable nits, lines user did not modify
->
-> **Output format**: structured report with severity-classified findings (critical / high / medium (only findings ≥80 confidence are reported)). One finding per line: `[severity] file:line — problem (one sentence); fix: suggested change`. If no findings, say so explicitly. No preamble, no end-of-report summary.
-
-(Role prompt is self-contained — works even when the upstream `code-review` user-skill / plugin isn't installed.)
-
-(Sister `~/.claude/commands/verify-code-review.md` is also generated by `harnessed setup` so `/verify-code-review` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed verify-code-review --apply` CLI claims are removed; that subcommand was never implemented.)
+Use the Bash tool to run:
+
+```bash
+echo "$ARGUMENTS" | harnessed run verify-code-review --task-stdin
+```
+
+If `$ARGUMENTS` is empty, run `harnessed run verify-code-review` (no stdin pipe).
+
+After completion, the Bash output prints a `Next:` hint on stderr suggesting the next stage. Decide whether to invoke based on conversation context — the hint is informational, not prescriptive.
+
+<!-- harnessed-generated:v3.4.4 -->
 
 ## References
 
 - D-04 Stage ④ Verify 7 sub 分解
-- ~/.claude/CLAUDE.md "Verify 阶段 — code-review 多 agent 并行" verbatim
-- ~/.claude/CLAUDE.md "子任务并行执行机制 — subagent vs Agent Teams 路由" subagent default
 - workflows/capabilities.yaml — code-review
 - workflows/judgments/parallelism-gate.yaml — subagent-default.fires
 - workflows/defaults.yaml — ralph_max_iterations.verify-code-review.* values (W2.2 backfill)

package/workflows/verify/code-review/workflow.yaml

@@ -1,12 +1,10 @@
 # workflows/verify/code-review/workflow.yaml — Phase v3.0-3.4 W0 T3.4.W0.11
 #
 # Stage ④.b verify sub-workflow — code-review 多 agent 并行 fan-out 高置信度 finding
-# (subagent default per CLAUDE.md 子任务并行机制 — Task / Agent 工具 spawn 多任务 fan-out,
+# (subagent default per bundled parallelism gate — Task / Agent 工具 spawn 多任务 fan-out,
 # context 隔离, token 敏感)。
 #
 # Sister refs:
-#   - ~/.claude/CLAUDE.md "Verify 阶段" — "code-review 多 agent 并行" 章节
-#   - ~/.claude/CLAUDE.md "子任务并行执行机制 — subagent vs Agent Teams 路由" subagent default
 #   - workflows/judgments/parallelism-gate.yaml — subagent-default.fires 默认 fan-out
 #   - workflows/capabilities.yaml — code-review (Bucket 1 mattpocock 高频招式)
 #   - workflows/verify-work/workflow.yaml v2 SHIPPED phase 03-code-review-parallel sister pattern