harnessed 3.4.1 → 3.4.3

package/workflows/discuss/auto/SKILL.md

@@ -5,7 +5,7 @@ description: |
   战略层 / Phase 层 / 子任务层独立判断 gate, 可能 3 个全跑 / 1-2 个 / 全跳 + 透明声明。
   schema_version: harnessed.workflow.v3 with delegates_to (3 sub: strategic + phase + subtask, mode parallel)
   + disciplines_applied (6 default) + tools_available (planning-with-files)。
-  Triggered by harnessed CLI `harnessed discuss --topic <text>` or slash command `/discuss`
+  Triggered by slash command `/discuss`
   (bare per ADR 0030 namespace policy D-02 LOCK) after `harnessed setup`.
 trigger_phrases:
   - "discuss"
@@ -49,9 +49,17 @@ Sister `workflows/capabilities.yaml`:
 
 ## Invocation
 
-- CLI: `harnessed discuss --topic "<text>"`
 - Slash command: `/discuss <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.
+
+**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>".
+
+(Sister `~/.claude/commands/discuss.md` is also generated by `harnessed setup` so `/discuss` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed discuss --apply` CLI claims are removed; that subcommand was never implemented.)
+
 ## References
 
 - D-01 master orchestrator delegation pattern

package/workflows/discuss/phase/SKILL.md

@@ -41,7 +41,6 @@ Sister `workflows/judgments/phase-gate.yaml`:
 
 ## Invocation
 
-- CLI: `harnessed discuss-phase --phase <num>`
 - Slash command: `/discuss-phase <num>` (after `harnessed setup`)
 
 ## Routing rules
@@ -52,13 +51,37 @@ Sister `workflows/judgments/phase-gate.yaml`:
 - < 1 天工作量
 - bug 修复且已有最小复现
 
+<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-Use the SlashCommand tool to run: `{{ capabilities.gsd-discuss-phase.cmd }}`
-
-(If the rendered cmd above is the bare `/gsd-discuss-phase` accompanied by a `⚠️ ... not installed`
-warning from `harnessed setup`, install the missing plugin first then re-run
-`harnessed setup` to re-render this SKILL.md with the full namespaced cmd.)
+**Preferred path** (when the upstream specialist is installed): use the SlashCommand tool to run `{{ capabilities.gsd-discuss-phase.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 **Phase clarification analyst**.
+>
+> **Mission**: Surface and resolve gray-area implementation decisions BEFORE a phase enters planning. Fires when ≥2 open decisions, cross-phase data flow is unclear, or scope spans >1 day. Adapted from GSD `/gsd-discuss-phase`.
+>
+> **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. List every open decision as a single question (1 line each)
+>
+> 2. For each, list 2-4 candidate answers with one-line tradeoffs
+>
+> 3. Identify cross-phase contracts (data flow / API shape / migration order)
+>
+> 4. Flag decisions blocking start (must answer before plan) vs. deferrable
+>
+> 5. Persist to `.planning/<phase>/findings.md` + `knowledge.md` for hand-off
+>
+> 6. If the layer is genuinely clear, say 'no clarification needed' and exit
+>
+> **Output format**: structured report with severity-classified findings (blocking / deferrable / 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 `gsd-discuss-phase` user-skill / plugin isn't installed.)
+
+(Sister `~/.claude/commands/discuss-phase.md` is also generated by `harnessed setup` so `/discuss-phase` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed discuss-phase --apply` CLI claims are removed; that subcommand was never implemented.)
 
 ## References
 

package/workflows/discuss/strategic/SKILL.md

@@ -6,7 +6,7 @@ description: |
   schema_version: harnessed.workflow.v3 with disciplines_applied (6 default) + tools_available
   (gstack-office-hours + gstack-plan-ceo-review + planning-with-files) + 3 phases (gate
   judgments.strategic-gate.* fires + planning-with-files /plan findings.md 持久化)。
-  Triggered by harnessed CLI `harnessed discuss-strategic --topic <text>` or slash command
+  Triggered by slash command
   `/discuss-strategic` after `harnessed setup`.
 trigger_phrases:
   - "discuss strategic"
@@ -45,7 +45,6 @@ Sister `workflows/judgments/strategic-gate.yaml`:
 
 ## Invocation
 
-- CLI: `harnessed discuss-strategic --topic "<text>"`
 - Slash command: `/discuss-strategic <text>` (after `harnessed setup`)
 
 ## Routing rules
@@ -56,13 +55,39 @@ Sister `workflows/judgments/strategic-gate.yaml`:
 - continuing 已有 phase 的执行
 - 用户已给明确 ticket / spec
 
+<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-Use the SlashCommand tool to run: `{{ capabilities.gstack-office-hours.cmd }}`
-
-(If the rendered cmd above is the bare `/gstack-office-hours` accompanied by a `⚠️ ... not installed`
-warning from `harnessed setup`, install the missing plugin first then re-run
-`harnessed setup` to re-render this SKILL.md with the full namespaced cmd.)
+**Preferred path** (when the upstream specialist is installed): use the SlashCommand tool to run `{{ capabilities.gstack-office-hours.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 **Strategic Office-Hours advisor (CEO + Product lens)**.
+>
+> **Mission**: Stress-test the product / scope / business value of a new feature, milestone, or project BEFORE engineering investment. Adapted from gstack `/office-hours` + `/plan-ceo-review`.
+>
+> **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. What user problem does this solve? Who specifically experiences it today?
+>
+> 2. Why this, why now? (alternative cost of working on something else)
+>
+> 3. What does success look like — measurable, not vibes (1 metric, not 5)?
+>
+> 4. Is the scope MVP-able? What's the smallest cut that still proves the bet?
+>
+> 5. What assumptions are load-bearing? Which would kill the feature if wrong?
+>
+> 6. Who pays the maintenance cost after ship — same team, or a hand-off?
+>
+> 7. Decision: ship / iterate / kill / table — with one-line reason
+>
+> **Output format**: structured report with severity-classified findings (ship / iterate / kill / table (each with reason)). 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 `gstack-office-hours` user-skill / plugin isn't installed.)
+
+(Sister `~/.claude/commands/discuss-strategic.md` is also generated by `harnessed setup` so `/discuss-strategic` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed discuss-strategic --apply` CLI claims are removed; that subcommand was never implemented.)
 
 ## References
 

package/workflows/discuss/subtask/SKILL.md

@@ -46,7 +46,6 @@ Sister `workflows/judgments/subtask-gate.yaml`:
 
 ## Invocation
 
-- CLI: `harnessed discuss-subtask --task "<text>"`
 - Slash command: `/discuss-subtask <text>` (after `harnessed setup`)
 
 ## Routing rules
@@ -57,13 +56,37 @@ Sister `workflows/judgments/subtask-gate.yaml`:
 - 标准库直接调用
 - bug 修复且已知根因
 
+<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-Use the SlashCommand tool to run: `{{ capabilities.superpowers-brainstorming.cmd }}`
-
-(If the rendered cmd above is the bare `/superpowers-brainstorming` accompanied by a `⚠️ ... not installed`
-warning from `harnessed setup`, install the missing plugin first then re-run
-`harnessed setup` to re-render this SKILL.md with the full namespaced cmd.)
+**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 brainstormer**.
+>
+> **Mission**: Generate ≥2 implementation approaches for a single subtask and compare tradeoffs. Fires when core algorithm / data structure / API contract / high error-cost. Skip pure CRUD or single-obvious-path tasks.
+>
+> **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. State the subtask in one sentence; confirm scope with user if ambiguous
+>
+> 2. Produce 2-4 distinct approaches (not just '2 flavors of the same idea')
+>
+> 3. For each: complexity, perf, failure modes, test surface, future change cost
+>
+> 4. Recommend one with 1-2 line reason; flag risks of the chosen path
+>
+> 5. Output a `findings.md` block the implementer can paste into the task
+>
+> 6. If options collapse to one (others clearly bad), say so and exit fast
+>
+> **Output format**: structured report with severity-classified findings (recommended / acceptable / rejected). 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/discuss-subtask.md` is also generated by `harnessed setup` so `/discuss-subtask` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed discuss-subtask --apply` CLI claims are removed; that subcommand was never implemented.)
 
 ## References
 

package/workflows/plan/architecture/SKILL.md

@@ -42,7 +42,6 @@ trigger `is-complex-architecture` 重构 ref。
 
 ## Invocation
 
-- CLI: `harnessed plan-architecture --module "<name>"`
 - Slash command: `/plan-architecture <name>` (after `harnessed setup`)
 
 ## Routing rules
@@ -53,13 +52,39 @@ trigger `is-complex-architecture` 重构 ref。
 - 性能 / scaling 关键路径
 - 引入显著技术债 / migration 风险
 
+<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-Use the SlashCommand tool to run: `{{ capabilities.plan-eng-review.cmd }}`
-
-(If the rendered cmd above is the bare `/plan-eng-review` accompanied by a `⚠️ ... not installed`
-warning from `harnessed setup`, install the missing plugin first then re-run
-`harnessed setup` to re-render this SKILL.md with the full namespaced cmd.)
+**Preferred path** (when the upstream specialist is installed): use the SlashCommand tool to run `{{ capabilities.plan-eng-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 **Staff Engineer architect**.
+>
+> **Mission**: Lock down system architecture BEFORE phase planning when complex (≥3 modules / new framework / new data model / scaling-critical / large migration). Adapted from gstack `/plan-eng-review`.
+>
+> **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. Identify the smallest architecture change that satisfies all requirements
+>
+> 2. Diagram component boundaries (data flow / call direction / ownership)
+>
+> 3. List interfaces / contracts between components (function signatures, API shapes)
+>
+> 4. Failure modes: what happens when each component is slow / down / inconsistent?
+>
+> 5. Migration / rollback path — can we ship in slices, or all-at-once?
+>
+> 6. Choose mechanisms with the lowest blast radius and lowest unique vocabulary
+>
+> 7. Document tradeoffs of the rejected alternatives (so reviewers see the road not taken)
+>
+> **Output format**: structured report with severity-classified findings (approved / approved-with-changes / 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 `plan-eng-review` user-skill / plugin isn't installed.)
+
+(Sister `~/.claude/commands/plan-architecture.md` is also generated by `harnessed setup` so `/plan-architecture` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed plan-architecture --apply` CLI claims are removed; that subcommand was never implemented.)
 
 ## References
 

package/workflows/plan/auto/SKILL.md

@@ -6,7 +6,7 @@ description: |
   task_plan.md。schema_version: harnessed.workflow.v3 with delegates_to (2 sub: architecture
   serial order 1 + phase serial order 2) + disciplines_applied (6 default) + tools_available
   (planning-with-files + plan-eng-review + gsd-plan-phase)。
-  Triggered by harnessed CLI `harnessed plan --topic <text>` or slash command `/plan`
+  Triggered by slash command `/plan`
   (bare per ADR 0030 namespace policy D-02 LOCK) after `harnessed setup`.
 trigger_phrases:
   - "plan"
@@ -50,9 +50,17 @@ Sister `workflows/capabilities.yaml`:
 
 ## Invocation
 
-- CLI: `harnessed plan --topic "<text>"`
 - Slash command: `/plan <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.
+
+**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>".
+
+(Sister `~/.claude/commands/plan.md` is also generated by `harnessed setup` so `/plan` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed plan --apply` CLI claims are removed; that subcommand was never implemented.)
+
 ## References
 
 - D-01 master orchestrator delegation pattern

package/workflows/plan/phase/SKILL.md

@@ -46,7 +46,6 @@ reframe)。
 
 ## Invocation
 
-- CLI: `harnessed plan-phase --phase <num>`
 - Slash command: `/plan-phase <num>` (after `harnessed setup`)
 
 ## Output artifacts
@@ -55,13 +54,39 @@ reframe)。
 - `progress.md` — phase 进度跟踪 + cross-session 恢复
 - (`findings.md` 由 discuss-* sub-workflow 产出, 此处不重复)
 
+<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-Use the SlashCommand tool to run: `{{ capabilities.gsd-plan-phase.cmd }}`
-
-(If the rendered cmd above is the bare `/gsd-plan-phase` accompanied by a `⚠️ ... not installed`
-warning from `harnessed setup`, install the missing plugin first then re-run
-`harnessed setup` to re-render this SKILL.md with the full namespaced cmd.)
+**Preferred path** (when the upstream specialist is installed): use the SlashCommand tool to run `{{ capabilities.gsd-plan-phase.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 **Phase planner**.
+>
+> **Mission**: Break a phase into ordered, dependency-aware tasks with explicit file paths and acceptance criteria, then persist via planning-with-files plugin. Adapted from GSD `/gsd-plan-phase` (Wave A research → Wave B planner → Wave C plan-checker).
+>
+> **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. Each task names the exact files it touches (NOT just 'auth module')
+>
+> 2. Each task has acceptance criteria a third party can verify
+>
+> 3. Dependencies are explicit (task N requires task M output)
+>
+> 4. Tasks are ≤1 day each; split if larger
+>
+> 5. Identify the verification step (test / lint / typecheck) for each task
+>
+> 6. Persist as `task_plan.md` + `progress.md` via planning-with-files `/plan`
+>
+> 7. Final pass: a fresh agent should be able to execute from these files alone
+>
+> **Output format**: structured report with severity-classified findings (ready-to-execute / needs-revision / 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 `gsd-plan-phase` user-skill / plugin isn't installed.)
+
+(Sister `~/.claude/commands/plan-phase.md` is also generated by `harnessed setup` so `/plan-phase` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed plan-phase --apply` CLI claims are removed; that subcommand was never implemented.)
 
 ## References
 

package/workflows/research/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: research
-description: 多源调研 workflow — Tavily/Exa/ctx7 多源 fan-out + GSD discuss synth aggregate; harnessed v2.0 NEW per R20.7 (Stage ① Discuss 独立 call); schema bumped to harnessed.workflow.v3 in Phase v3.0-3.4 W1.1 (T3.4.W1.1) with disciplines_applied [6] + tools_available [tavily-mcp, exa-mcp, ctx7, gsd-discuss-phase], phases reuse v2 verbatim. Triggered by harnessed CLI `harnessed research --topic <text>` or slash command `/research` after `harnessed setup`.
+description: 多源调研 workflow — Tavily/Exa/ctx7 多源 fan-out + GSD discuss synth aggregate; harnessed v2.0 NEW per R20.7 (Stage ① Discuss 独立 call); schema bumped to harnessed.workflow.v3 in Phase v3.0-3.4 W1.1 (T3.4.W1.1) with disciplines_applied [6] + tools_available [tavily-mcp, exa-mcp, ctx7, gsd-discuss-phase], phases reuse v2 verbatim. Triggered by slash command `/research` after `harnessed setup`.
 preamble-tier: 2
 schema_version: harnessed.workflow.v3
 ---
@@ -28,7 +28,6 @@ Sister `workflows/capabilities.yaml` entries:
 - `gsd-discuss-phase` (synth aggregate)
 
 ## Invocation
-- CLI: `harnessed research --topic "<topic>"`
 - Slash command: `/research <topic>` (after `harnessed setup`)
 
 ## Routing rules (sister ~/.claude/rules/web-search.md)
@@ -37,3 +36,33 @@ Sister `workflows/capabilities.yaml` entries:
 - 库 / API 文档 → ctx7 CLI (per ~/.claude/rules/context7.md)
 - 关键词 / 时效内容 → Tavily MCP (默认)
 - 抓整站 / 站点结构 → Tavily crawl/map
+
+<!-- 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.
+
+**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 **Research analyst**.
+>
+> **Mission**: Multi-source investigation (docs / web search / codebase grep / library probe) producing a `findings.md` with citations, NOT speculation. Use `ctx7` for library docs, `tavily-mcp` / `exa-mcp` for web, `gh` CLI for GitHub artifacts, and codebase `Grep` for internal references.
+>
+> **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. Resolve each unknown claim to a citable source (URL, file:line, or `ctx7` doc id)
+>
+> 2. Cite version explicitly when discussing library / framework APIs (training cutoff may be stale)
+>
+> 3. Capture conflicting sources side-by-side; do not silently pick one
+>
+> 4. Flag `OPEN: <question>` for items the user must decide; never paper over
+>
+> 5. Persist results to `.planning/<phase>/findings.md` for cross-session handoff
+>
+> **Output format**: structured report with severity-classified findings (verified / unverified / conflicting / open). 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 `specialist` user-skill / plugin isn't installed.)
+
+(Sister `~/.claude/commands/research.md` is also generated by `harnessed setup` so `/research` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed research --apply` CLI claims are removed; that subcommand was never implemented.)

package/workflows/retro/SKILL.md

@@ -47,23 +47,37 @@ Sister `workflows/capabilities.yaml` entries:
 - ✅ **触发**: 项目结束 / 里程碑结束 / 用户明示 "复盘 / retro / lessons learned"
 - ❌ **跳过**: 日常 PR / 单 phase 完成 (常规 verify-progress 已够用)
 
-## CLI invocation
-
-```bash
-# Dry-run preview — arbitrate-only, never spawns SDK.
-harnessed retro --milestone <name> --dry-run --non-interactive
-
-# Apply path — real SDK spawn + 2-phase serial (gstack /retro → planning-with-files RETROSPECTIVE.md sink).
-harnessed retro --milestone <name> --apply
-```
-
+<!-- v3.4.3-dual-path-invocation -->
 ## How to invoke
 
-Use the SlashCommand tool to run: `{{ capabilities.retro-gstack.cmd }}`
-
-(If the rendered cmd above is the bare `/retro-gstack` accompanied by a `⚠️ ... not installed`
-warning from `harnessed setup`, install the missing plugin first then re-run
-`harnessed setup` to re-render this SKILL.md with the full namespaced cmd.)
+**Preferred path** (when the upstream specialist is installed): use the SlashCommand tool to run `{{ capabilities.retro-gstack.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 **Retrospective facilitator**.
+>
+> **Mission**: Run a Lessons / Decisions / Surprises retrospective for the closed milestone, then persist to `RETROSPECTIVE.md`. Adapt the gstack `/retro` method when available; otherwise structure the conversation yourself.
+>
+> **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. What did we set out to do, vs. what actually shipped?
+>
+> 2. Top 3 surprises (positive or negative) — root cause each
+>
+> 3. Decisions that paid off; decisions we would reverse
+>
+> 4. Process changes for next milestone (concrete, not vague)
+>
+> 5. What deserves a permanent rule entry (CLAUDE.md / docs/adr/)?
+>
+> 6. Persist verbatim to `.planning/RETROSPECTIVE.md` — append, do not overwrite
+>
+> **Output format**: structured report with severity-classified findings (lesson / decision / surprise / process-change). 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 `retro-gstack` user-skill / plugin isn't installed.)
+
+(Sister `~/.claude/commands/retro.md` is also generated by `harnessed setup` so `/retro` is a real platform slash command — both files carry the same dual-path instruction. Previous v3.4.x `harnessed retro --apply` CLI claims are removed; that subcommand was never implemented.)
 
 ## References