@exaudeus/workrail 0.8.6 → 0.9.0

package/workflows/coding-task-workflow-agentic.json

@@ -0,0 +1,262 @@
+{
+  "id": "coding-task-workflow-agentic",
+  "name": "Agentic Task Dev Workflow (Invariants • Architecture • Vertical Slices • PR Sizing • Subagent Audits)",
+  "version": "1.0.0",
+  "description": "A less-restrictive, reflection-heavy task development workflow aligned with WorkRail’s agentic philosophy: create invariants, capture architecture decisions, plan vertical slices, right-size PRs, iterate via audits (plan/context/validation), and implement slice-by-slice with checkpoints and resume support.",
+  "preconditions": [
+    "User provides a task description (ticket text or equivalent) and success criteria (even if partial).",
+    "Agent has read access to the codebase and can use tools (search/read/edit/terminal).",
+    "A validation path exists (tests, build, or a deterministic verification strategy).",
+    "If the task touches production or critical paths, rollback/flag strategy can be defined."
+  ],
+  "clarificationPrompts": [
+    "What are the acceptance criteria and explicit non-goals?",
+    "What are the key invariants? (backwards compatibility, API contracts, data correctness, performance budgets, security constraints)",
+    "Any rollout constraints? (feature flags, staged rollout, migrations, telemetry requirements)",
+    "Any constraints on tooling? (can/can’t run tests locally, CI only, limited environment access)",
+    "Any preferred code patterns/examples in this repo that should be followed?"
+  ],
+  "metaGuidance": [
+    "DEFAULT BEHAVIOR: self-execute with tools. Only ask the user for (a) business decisions, (b) missing external artifacts, or (c) permissions/constraints you cannot resolve.",
+    "ARCHITECTURE OVER PATCHES: prefer structural fixes that eliminate whole classes of issues; avoid local patches unless they are explicitly intended.",
+    "INVARIANTS FIRST: define constraints before designing or coding; use invariants to evaluate plan/slices/PR size.",
+    "CONTEXT DOC: For non-small tasks, create a base CONTEXT.md early (chat-first; write the file if possible). Never commit documentation files unless the user explicitly asks.",
+    "USER RULES: Capture user rules/preferences early as `userRules` (in CONTEXT.md). When ideating and planning, explicitly apply `userRules` and call out deviations with rationale + approval request.",
+    "VERTICAL SLICES: plan and implement as independently testable, reviewable slices; each slice should be mergeable and ideally shippable behind a flag.",
+    "PR/MR SIZING: avoid monster PRs; if scope is large, split into multiple PRs by slices with explicit dependencies.",
+    "REFLECTION LOOPS: iterate on context and plan with bounded loops; audit yourself before coding and before finalizing.",
+    "SUBAGENTS: delegate ONLY when a step explicitly instructs delegation; prefer auditor-style delegation (challenge/audit/validate) over executor-style delegation.",
+    "VALIDATION: prefer compile-time safety and deterministic tests; verify each slice before moving on; fail fast with meaningful errors."
+  ],
+  "steps": [
+    {
+      "id": "phase-0-triage-and-mode",
+      "title": "Phase 0: Triage (Complexity • Risk • Automation • Doc Depth • PR Strategy)",
+      "prompt": "**ANALYZE** the task and classify with deterministic criteria.\n\n## 0) Rigor mode (deterministic)\nSelect **rigorMode**: QUICK / STANDARD / THOROUGH.\n\nScore each criterion 0–2 and sum. Use the table:\n- **Scope breadth** (files/areas touched): 0=1–2 files, 1=multi-file but single area, 2=multi-area\n- **Risk level**: 0=low, 1=moderate, 2=high (security/auth/data loss/release pipeline/perf critical)\n- **Uncertainty**: 0=clear requirements + known code path, 1=some ambiguity, 2=unknowns/missing acceptance criteria\n- **Repro difficulty**: 0=deterministic + local, 1=some async/edge cases, 2=flaky/CI-only/racy\n- **Externalities**: 0=internal-only, 1=some external deps, 2=publishing/infra/3rd-party integration\n\nDecision:\n- 0–2 → **QUICK**\n- 3–5 → **STANDARD**\n- 6–10 → **THOROUGH**\n\nAlso set:\n- QUICK: `auditDepth=light`, `maxQuestions=1`, `maxParallelism=0`\n- STANDARD: `auditDepth=normal`, `maxQuestions=3`, `maxParallelism=1`\n- THOROUGH: `auditDepth=deep`, `maxQuestions=5`, `maxParallelism=3`\n\n## 1) taskComplexity\nSmall / Medium / Large\n- Small: 1–2 files, low risk, clear change, minimal ambiguity\n- Medium: multi-file, moderate risk, some ambiguity, needs planning\n- Large: architectural impact, multiple systems, high risk/unknowns\n\n## 2) riskLevel\nLow / Medium / High\n- High if: auth/payments/security/data integrity/perf-sensitive/production incident/release pipeline\n\n## 3) automationLevel\nHigh / Medium / Low\n- High: proceed autonomously; ask only for real decisions\n- Medium: normal confirmations at gates\n- Low: extra confirmations and explicit checklists\n\n## 4) docDepth (chat artifacts only; no auto-committed docs)\nNone / Light / Full\n- None: concise plan artifact only\n- Light: plan artifact + checkpoint/resume summary\n- Full: plan artifact + risk register + verification matrix + checkpoint/resume summary\n\n## 5) prStrategy\nSinglePR / MultiPR\n- MultiPR if Large or diff is broad (many files/domains)\n\n**Set context variables**: `rigorMode`, `auditDepth`, `maxQuestions`, `maxParallelism`, `taskComplexity`, `riskLevel`, `automationLevel`, `docDepth`, `prStrategy`.\n\n**VERIFY (minimal questions)**: ask the user to confirm or override `rigorMode` and `prStrategy` only if it impacts delivery expectations.",
+      "requireConfirmation": true
+    },
+    {
+      "id": "phase-0b-minimum-context-request",
+      "title": "Phase 0b: Minimum Inputs Gate (Only Ask What You Truly Need)",
+      "prompt": "If any critical information is missing, request ONLY the minimum needed to proceed.\n\n**Ask for:**\n- Ticket text / requirements (if not provided)\n- Success criteria / expected behavior\n- Constraints (permissions, environment, deadlines)\n- Pointers to relevant code areas (if user has them)\n\n**Do NOT ask** questions you can answer via tools.\n\n**Output:** a short list of missing inputs (if any) and proceed once answered.",
+      "requireConfirmation": {
+        "or": [
+          { "var": "automationLevel", "equals": "Low" },
+          { "var": "automationLevel", "equals": "Medium" }
+        ]
+      }
+    },
+    {
+      "id": "phase-0c-base-context-doc",
+      "title": "Phase 0c: Base Context Doc (Non-Small)",
+      "runCondition": { "var": "taskComplexity", "not_equals": "Small" },
+      "prompt": "Create a base `CONTEXT.md` as the durable artifact for this workflow run.\n\n**Rules:**\n- Default deliverable is **chat-first** (pasteable content).\n- If the agent can write files, also write/update `CONTEXT.md`.\n- Do NOT commit documentation files unless the user explicitly asks.\n\n**CONTEXT.md structure (must include):**\n1) **Task Summary** (1 paragraph)\n2) **Triage**: rigorMode, taskComplexity, riskLevel, automationLevel, docDepth, prStrategy\n3) **Inputs & Sources**: ticket links/text pointers; any user-provided file paths\n4) **User Rules & Preferences (`userRules`)**:\n   - Extract relevant user rules/preferences from user instructions + repo docs/patterns.\n   - Keep this focused and actionable.\n   - Set context variable `userRules` as a bulleted list.\n5) **Open Questions**: only those that require user decisions (cap at `maxQuestions`)\n6) **Assumptions**: explicitly labeled\n7) **Placeholders**:\n   - Invariants\n   - Architecture decision\n   - Slice/PR plan\n   - Plan findings / follow-up tickets\n   - Verification notes\n8) **Resumption Instructions**:\n   - `workflow_get` id: `coding-task-workflow-agentic`\n   - `workflow_next` with completedSteps + key context vars\n\n**Output:** the full content for `CONTEXT.md` (or confirm file written).",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-1-context-gathering",
+      "title": "Phase 1: Context Gathering (Main Agent, Tool-Driven)",
+      "runCondition": { "var": "taskComplexity", "not_equals": "Small" },
+      "prompt": "Gather enough context to design and plan correctly.\n\n**Rules:**\n- Do this yourself (no delegation in this step).\n- Use tools to verify everything.\n- Prefer matching existing patterns over inventing new ones.\n- Prefer answering your own questions with tools; only keep true human-decision questions.\n\n**Deliverable (in chat, concise):**\n- Entry points and call chain sketch (file references)\n- Key modules/classes/functions involved\n- Existing patterns that apply (with 2–3 concrete examples)\n- Testing approach found in repo (where tests live; key helpers)\n- Risks/unknowns list\n\n**Question resolution pass (required):**\n- For uncertainties you encounter, attempt resolution via tools/code first.\n- Only add to `openQuestions` if it is a true business/product decision.\n- Enforce: `openQuestions.length <= maxQuestions`.\n\n**Set context variables:**\n- `contextSummary` (short)\n- `candidateFiles` (list of key file paths)\n- `openQuestions` (true human decisions only)\n\nAlso keep `userRules` in mind: if you discover a conflict between repo patterns and `userRules`, note it explicitly for planning."
+    },
+    {
+      "id": "phase-1b-context-audit-mode-adaptive",
+      "title": "Phase 1b: Context Audit (Mode-Adaptive, Subagent-Friendly)",
+      "runCondition": {
+        "or": [
+          { "var": "rigorMode", "equals": "THOROUGH" },
+          {
+            "and": [
+              { "var": "rigorMode", "equals": "STANDARD" },
+              { "var": "riskLevel", "equals": "High" }
+            ]
+          }
+        ]
+      },
+      "prompt": "Audit your context understanding before designing.\n\nMode behavior:\n- **QUICK**: skip this step (should not run)\n- **STANDARD**: do a self-audit; delegate at most once if you have subagent support\n- **THOROUGH**: delegate in parallel if you have subagent support\n\n**If subagent support is available and `rigorMode=THOROUGH`:**\n- Delegate to the WorkRail Executor **TWICE SIMULTANEOUSLY** using the **Context Gathering Routine** in **audit mode**:\n  1) Focus: COMPLETENESS (what did I miss?)\n  2) Focus: DEPTH (what did I misunderstand or stay too shallow on?)\n\n**If `rigorMode=STANDARD`:**\n- Prefer self-audit. Optionally delegate ONCE focusing on COMPLETENESS.\n\n**If no subagents:** do a self-audit using the two lenses.\n\n**SYNTHESIZE** audit findings:\n- Update `contextSummary` with gaps filled\n- Update `openQuestions` but keep it <= `maxQuestions` and only for true user decisions\n\n**Quality gate:** proceed only if you can explain the relevant flow end-to-end.",
+      "prompt": "Audit your context understanding before designing.\n\nMode behavior:\n- **QUICK**: skip this step (should not run)\n- **STANDARD**: do a self-audit; delegate at most once if you have subagent support\n- **THOROUGH**: delegate in parallel if you have subagent support\n\n**If subagent support is available and `rigorMode=THOROUGH`:**\n- Delegate to the WorkRail Executor **TWICE SIMULTANEOUSLY** using the **Context Gathering Routine** in **audit mode**:\n  1) Focus: COMPLETENESS (what did I miss?)\n  2) Focus: DEPTH (what did I misunderstand or stay too shallow on?)\n\n**If `rigorMode=STANDARD`:**\n- Prefer self-audit. Optionally delegate ONCE focusing on COMPLETENESS.\n\n**If no subagents:** do a self-audit using the two lenses.\n\n**SYNTHESIZE** audit findings:\n- Update `contextSummary` with gaps filled\n- Resolve uncertainties with tools when possible\n- Update `openQuestions` but keep it <= `maxQuestions` and only for true human decisions\n\n**Quality gate:** proceed only if you can explain the relevant flow end-to-end.",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-2-invariants-and-nongoals",
+      "title": "Phase 2: Invariants (Contracts, Constraints, Non-Goals)",
+      "prompt": "Create explicit invariants and non-goals.\n\n**Include (as applicable):**\n- API/behavior contracts that must not change\n- Data invariants (schema constraints, idempotency, ordering)\n- Performance budgets (latency, allocations, query counts)\n- Security/privacy constraints\n- Rollout invariants (flagging, migration safety, rollback)\n- Non-goals (explicitly out of scope)\n\n**Output:** a numbered list of invariants + non-goals.\n\n**Set context variables:** `invariants`, `nonGoals`.\n\n**VERIFY:** ask the user to confirm only if any invariant is a product decision.",
+      "requireConfirmation": {
+        "or": [
+          { "var": "automationLevel", "equals": "Low" },
+          { "var": "riskLevel", "equals": "High" }
+        ]
+      }
+    },
+    {
+      "id": "phase-3-architecture-capture",
+      "title": "Phase 3: Architecture Capture (Options → Recommendation → Risks)",
+      "runCondition": { "var": "taskComplexity", "not_equals": "Small" },
+      "prompt": "Propose architecture options and choose a recommendation.\n\n**Do:**\n1) Identify 2–3 viable approaches (including a conservative one)\n2) Evaluate trade-offs against `invariants`\n3) Recommend one approach\n4) Identify risks and mitigations\n5) Identify what must be proven early (spikes / thin slices)\n\n**If `rigorMode=THOROUGH` and subagents are available:**\n- Delegate to the WorkRail Executor for **parallel ideation** (2–3 perspectives) using **Ideation Routine** (e.g., simplicity / maintainability / performance).\n\n**Output:**\n- Selected approach + rationale\n- “Architecture delta” (what changes structurally)\n- Risks + mitigations\n\n**Set context variables:** `architectureDecision`, `architectureRisks`.",
+      "requireConfirmation": {
+        "or": [
+          { "var": "automationLevel", "equals": "Low" },
+          { "var": "taskComplexity", "equals": "Large" }
+        ]
+      }
+    },
+    {
+      "id": "phase-4-vertical-slices-and-pr-plan",
+      "title": "Phase 4: Vertical Slices + PR Plan (Avoid Monster PRs)",
+      "prompt": "Create a vertical-slice plan and PR strategy.\n\n**Define `slices`** as an ordered list where each slice is:\n- Independently testable\n- Reviewable within a small PR\n- Compatible with `invariants`\n\nEach slice must include:\n- Name\n- Scope summary\n- Key files/components\n- Verification plan (tests/build/manual)\n- Rollout considerations (if any)\n\n**PR sizing gate:**\n- If `prStrategy = MultiPR`, map slices to PRs.\n- If `prStrategy = SinglePR` but the plan implies broad changes, recommend switching to MultiPR.\n\n**Set context variables:**\n- `slices` (array)\n- `estimatedPRCount` (number)\n- `prStrategyRationale` (short)\n\n**VERIFY:** user confirms PR strategy if it affects delivery expectations.",
+      "requireConfirmation": true
+    },
+    {
+      "id": "phase-4b-locks-compliance-audit",
+      "title": "Phase 4b: Locks Compliance Audit (Canonical Docs → Slices Matrix)",
+      "runCondition": {
+        "and": [
+          { "var": "taskComplexity", "not_equals": "Small" },
+          { "var": "riskLevel", "equals": "High" }
+        ]
+      },
+      "prompt": "Audit your slice plan against canonical lock documents to ensure no locked requirements are accidentally omitted.\n\n**This step exists to prevent the planning gap we just encountered (abstracting \"projections\" without enumerating all locked event schemas + idempotency enforcement).**\n\n**Do:**\n\n1) **Identify canonical docs** referenced in context/invariants (e.g., ADRs, `*-locks.md`, contract specs).\n   - If none exist, skip this audit (no canonical locks to check).\n\n2) **Extract locked items** from those docs:\n   - Search for: \"MUST\", \"locked\", \"required\", \"invariant\", closed-set enums, explicit schema lists, enforcement rules.\n   - List each as a distinct **lock ID** with a short description.\n\n3) **Map locks to slices** (create a matrix table):\n   - For each lock, identify which slice covers it.\n   - If a lock is not covered, either:\n     - add it to an existing slice (update `slices`), or\n     - add a new slice, or\n     - mark \"Deferred to [next phase/PR]\" with explicit reasoning.\n\n4) **Output the locks-to-slices matrix** (table format):\n\n| Lock ID | Lock Description | Mapped Slice | Status | Notes |\n|---------|------------------|--------------|--------|-------|\n| L1      | Idempotency via dedupeKey | Slice 2 | ✅ Mapped | enforced in append() |\n| L2      | Event schema closed set (12 kinds) | Multiple slices | ✅ Mapped | incremental |\n| L3      | Preferred tip last-activity | Projections | ⚠️ GAP | add to run DAG slice |\n\n**Set context variable:**\n- `locksMatrix` (the table for later reference)\n\n**VERIFY (if gaps detected):**\n- Require user confirmation to either: update slices to cover gaps, or explicitly defer with reasoning.\n- Do not proceed with unmapped locks unless user approves the deferral.",
+      "requireConfirmation": {
+        "or": [
+          { "var": "automationLevel", "equals": "Low" }
+        ]
+      }
+    },
+    {
+      "id": "phase-5-plan-iteration-init",
+      "title": "Phase 5: Plan Iteration Init (Bounded Loop Setup)",
+      "runCondition": { "var": "taskComplexity", "not_equals": "Small" },
+      "prompt": "Initialize a bounded plan-iteration loop.\n\n**Set:** `continuePlanning = true`.\n\nRule: max 3 iterations. Stop early when audits find no Major risks/gaps.",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-5-plan-iterations",
+      "type": "loop",
+      "title": "Phase 5: Plan Iteration Loop (Draft → Audit → Refocus)",
+      "runCondition": { "var": "taskComplexity", "not_equals": "Small" },
+      "loop": {
+        "type": "while",
+        "condition": { "var": "continuePlanning", "equals": true },
+        "maxIterations": 3
+      },
+      "body": [
+        {
+          "id": "phase-5a-draft-implementation-plan",
+          "title": "Plan Artifact Draft/Update (Strict Schema)",
+          "prompt": "Create or update the **Plan Artifact**. This is a deterministic, WorkRail-friendly contract that reduces variance.\n\n**Output the Plan Artifact in chat** using these exact headings (keep concise, but complete):\n\n1) **Problem statement**\n2) **Acceptance criteria** (bullets)\n3) **Non-goals** (bullets)\n4) **User rules/preferences applied**\n   - Summarize relevant `userRules` and state how the plan respects them.\n   - If any rule cannot be followed, document rationale + mitigation + add a user decision to `openQuestions` (counts toward `maxQuestions`).\n5) **Invariants** (bullets; must reference `invariants`)\n6) **Proposed approach** (1–2 paragraphs)\n7) **Alternatives considered** (at least 1 for STANDARD/THOROUGH) + why rejected\n8) **Vertical slices** (must match `slices`; each slice must include: scope, done-definition, key files/components, verification)\n9) **Test plan** (unit/integration/e2e/manual; reference repo patterns found)\n10) **Risk register** (top risks + mitigation + rollback/flag if relevant)\n11) **PR packaging** (SinglePR vs MultiPR; if MultiPR, state the rule: “package a PR at the end of each slice”)\n\n**Set context variables:**\n- `planArtifact` (the structured artifact)\n- `implementationPlan` (slice-by-slice actionable steps derived from the artifact)\n\n**VERIFY:** the artifact is concrete enough that another engineer could implement it without guessing.",
+          "requireConfirmation": false
+        },
+        {
+          "id": "phase-5b-plan-audit-mode-adaptive",
+          "title": "Plan Audit (Subagent-Friendly)",
+          "prompt": "Audit the current `planArtifact` / `implementationPlan`.\n\nMode behavior:\n- **QUICK**: self-audit only (no delegation)\n- **STANDARD**: self-audit; delegate at most once if subagents exist\n- **THOROUGH**: parallel delegation if subagents exist\n\n**If subagents available and `rigorMode=THOROUGH`:**\n- Delegate **THREE TIMES SIMULTANEOUSLY** to WorkRail Executor:\n  1) **Plan Analysis Routine**: completeness/pattern adherence/risks\n  2) **Hypothesis Challenge Routine (rigor=3)**: try to find holes / missed edge cases\n  3) **Execution Simulation Routine**: simulate the riskiest slice and likely failure modes\n\n**If subagents available and `rigorMode=STANDARD`:**\n- Delegate **ONCE** using **Plan Analysis Routine** focusing on completeness.\n\n**Otherwise:** do a self-audit using the same three lenses.\n\n**Output:**\n- Findings grouped by severity: Critical / Major / Minor\n- Concrete plan amendments\n\n**Set context variables:**\n- `planFindings`\n- `planAmendments`\n- `planConfidence` (1–10)",
+          "requireConfirmation": false
+        },
+        {
+          "id": "phase-5c-refocus-and-ticket-extraction",
+          "title": "Refocus: Incorporate Findings + Extract Follow-Up Tickets",
+          "prompt": "Apply valid amendments, and refocus.\n\n**Do:**\n- Update `planArtifact` and `implementationPlan` to incorporate `planAmendments`.\n- Extract valuable-but-out-of-scope work into `followUpTickets`.\n- Ensure plan still follows `invariants` and stays slice-oriented.\n\n**Set context variables:**\n- `followUpTickets`\n\n**VERIFY:** the plan is coherent and still PR-sized by slice.",
+          "requireConfirmation": {
+            "or": [
+              { "var": "automationLevel", "equals": "Low" },
+              { "var": "planConfidence", "lt": 8 }
+            ]
+          }
+        },
+        {
+          "id": "phase-5d-loop-exit-decision",
+          "title": "Loop Exit Decision (Fail-Safe): Set continuePlanning",
+          "prompt": "This step is **non-optional** and exists to prevent the common failure mode of forgetting to set `continuePlanning`.\n\n**You MUST output exactly:**\n- `continuePlanning`: true|false\n- `reason`: one sentence\n- `whatChangedSinceLastIteration`: 1–3 bullets\n\n**DEFAULT RULE:** set `continuePlanning=false` unless one of these is true:\n1) Any **Critical** finding remains unresolved\n2) Any **Major** finding remains unresolved\n3) `planConfidence < 8`\n4) A user decision in `openQuestions` blocks correctness\n5) A high-risk invariant has no test/verification path\n\nIf you set `continuePlanning=true`, name which condition(s) triggered it and what you will change in the next iteration.\n\n**Set context variables:** `continuePlanning`.",
+          "requireConfirmation": false
+        }
+      ]
+    },
+    {
+      "id": "phase-5e-doc-artifacts",
+      "title": "Phase 5e: Plan Artifact Output (Mode-Dependent, Chat-First)",
+      "runCondition": {
+        "or": [
+          { "var": "docDepth", "equals": "Light" },
+          { "var": "docDepth", "equals": "Full" }
+        ]
+      },
+      "prompt": "Produce mode-appropriate, **chat-first** artifacts derived from `planArtifact`.\n\n- If `docDepth = Light`: output `planArtifact` + a short **Checkpoint/Resume** section (current state + next slice).\n- If `docDepth = Full`: output `planArtifact` + a concise **Risk register** + **Verification matrix** + **Checkpoint/Resume**.\n\n**Rule:** do not create or commit documentation files unless the user explicitly asks. The default deliverable is in-chat text that the user can save.\n\n**Save/resume support:** include a concise checkpoint section describing current state and next slice.",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-6-test-strategy-optional",
+      "title": "Phase 6 (Optional): Separate Test Strategy & Scaffolding",
+      "runCondition": {
+        "and": [
+          { "var": "taskComplexity", "not_equals": "Small" },
+          { "var": "riskLevel", "equals": "High" }
+        ]
+      },
+      "prompt": "Optionally do a test-first slice when risk is high.\n\n**Do:**\n- Define the minimal test harness needed to validate invariants.\n- Implement scaffolding/tests for the riskiest slice first (without finishing all features).\n- Ensure tests fail for the right reason before implementing the behavior.\n\n**VERIFY:** tests meaningfully protect the invariants and the intended behavior.",
+      "requireConfirmation": true
+    },
+    {
+      "id": "phase-7-implement-slices",
+      "type": "loop",
+      "title": "Phase 7: Implement Slice-by-Slice (PREP → IMPLEMENT → VERIFY → CHECKPOINT)",
+      "runCondition": {
+        "and": [
+          { "var": "taskComplexity", "not_equals": "Small" }
+        ]
+      },
+      "loop": {
+        "type": "forEach",
+        "items": "slices",
+        "itemVar": "currentSlice",
+        "indexVar": "sliceIndex",
+        "maxIterations": 20
+      },
+      "body": [
+        {
+          "id": "phase-7a-prep",
+          "title": "PREP: Prepare Slice {{sliceIndex}}",
+          "prompt": "Prepare to implement the slice `{{currentSlice.name}}`.\n\n**Do:**\n- Re-state slice goal and verification\n- Identify exact files/components to change\n- Re-check invariants impacted by this slice\n- Match existing patterns (find 1–3 relevant exemplars)\n- Apply `userRules` (if any rule affects this slice, call it out)\n\n**Pre-change sanity check (required, grouped):**\n- Existence check: key files/symbols referenced by the slice exist\n- Signature/contract check: confirm any key function/type signatures the slice assumes\n- Bounded impact scan: find 1–2 hop callers/callees for the primary touchpoints\n- Scope reconciliation: if you discover additional required touchpoints beyond the slice plan, update the slice scope (or stop and re-plan)\n\n**Stop and ask** only if a business decision is required or a prerequisite is missing.",
+          "requireConfirmation": false
+        },
+        {
+          "id": "phase-7b-implement",
+          "title": "IMPLEMENT: Slice {{sliceIndex}}",
+          "prompt": "Implement ONLY the current slice `{{currentSlice.name}}`.\n\n**Constraints:**\n- Keep the diff PR-sized\n- Prefer architectural moves that reduce future work\n- No drive-by refactors unless required by invariants\n\n**Self-execute** with tools; do not delegate implementation unless explicitly directed by the user.",
+          "requireConfirmation": false
+        },
+        {
+          "id": "phase-7c-verify",
+          "title": "VERIFY: Slice {{sliceIndex}}",
+          "prompt": "Verify the slice.\n\n**Do:**\n- Run the relevant tests/build\n- Add/adjust tests if needed to cover behavior\n- Ensure invariants still hold\n\n**If verification fails:** debug and fix deterministically; if blocked by environment constraints, request the user to run the necessary command and share output.",
+          "requireConfirmation": false
+        },
+        {
+          "id": "phase-7d-checkpoint",
+          "title": "CHECKPOINT: Save/Resume + PR Packaging Notes",
+          "prompt": "Checkpoint after slice completion.\n\n**Record:**\n- What changed (high level)\n- How it was verified\n- What remains (next slice)\n- Any follow-up tickets discovered\n- PR notes: if `prStrategy=MultiPR`, propose which slice(s) form the next PR.\n\nIf using docs, update the checkpoint/resume section.",
+          "requireConfirmation": false
+        },
+        {
+          "id": "phase-7e-multi-pr-packaging-gate",
+          "title": "PR Packaging Gate (Hard Stop when MultiPR)",
+          "runCondition": { "var": "prStrategy", "equals": "MultiPR" },
+          "prompt": "This is a **hard gate** to prevent PR size drift.\n\n**If `prStrategy=MultiPR`, you must stop here and package a PR before proceeding to the next slice.**\n\n**Output (PR-ready):**\n- Proposed PR title\n- 3–6 bullet summary (why, not what)\n- Test plan (what you ran)\n- Rollout/risks (if any)\n- What remains (next slice)\n\n**Then stop and wait for user confirmation** to proceed to the next slice.\n\n(Do not actually merge; do not assume permissions to push/create PR unless the user requests.)",
+          "requireConfirmation": true
+        }
+      ]
+    },
+    {
+      "id": "phase-8-small-task-fast-path",
+      "title": "Phase 8 (Small Only): Fast Path (Minimal Ceremony)",
+      "runCondition": { "var": "taskComplexity", "equals": "Small" },
+      "prompt": "For Small tasks, do:\n\n1) Confirm target locations with tools (existence + pattern match)\n2) Implement the smallest correct change\n3) Verify (tests/build or deterministic check)\n4) Provide a concise PR-ready summary\n\nAvoid heavy docs unless risk unexpectedly increases.",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-9-final-validation-and-handoff",
+      "title": "Phase 9: Final Validation + PR/MR Handoff (No Auto-Merge)",
+      "prompt": "Final validation and handoff.\n\n**Do:**\n- Verify acceptance criteria and invariants\n- Confirm test/build status and coverage gaps (if any)\n- Summarize slice completion and PR strategy outcome\n- Provide a PR/MR description draft (concise): summary + test plan + rollout notes\n- Provide follow-up tickets list (if any)\n\n**Important:** do not auto-merge, squash-merge, or delete branches as part of this workflow.",
+      "requireConfirmation": true
+    }
+  ]
+}

package/workflows/design-thinking-workflow-autonomous.agentic.json

@@ -0,0 +1,215 @@
+{
+  "id": "design-thinking-workflow-autonomous-agentic",
+  "name": "Design Thinking Workflow (Autonomous, Tiered Agent Cascade)",
+  "description": "Autonomous design thinking: minimal human input; doc-first execution; supports Agent Cascade Protocol tiers (Solo/Proxy/Delegation) with explicit fallbacks.",
+  "version": "0.1.0",
+  "preconditions": [
+    "User can provide a short problem/opportunity statement and constraints",
+    "Agent can create and update a single canonical design doc file for resumption"
+  ],
+  "clarificationPrompts": [
+    "What problem/opportunity are we addressing (1–3 paragraphs)?",
+    "What constraints matter (hard vs soft)?",
+    "What are the non-negotiable principles and anti-goals?",
+    "What is the intended deliverable (proposal, spec, roadmap, prototype API, etc.)?"
+  ],
+  "metaGuidance": [
+    "AUTONOMOUS: Minimize human input. If information is missing, create explicit assumptions and a validation plan; do not invent facts silently.",
+    "DOC-FIRST: `designDocPath` is the canonical source of truth. Keep chat output short; write the real work into the doc.",
+    "AGENT CASCADE PROTOCOL: Use the highest tier available and fall back explicitly:",
+    "- Tier 3 (Delegation / Gold): Subagents can call WorkRail MCP tools → delegate full routines/work packages.",
+    "- Tier 2 (Proxy / Silver): Subagents exist but cannot call WorkRail tools → delegate the work via natural language; you (main agent) write/synthesize.",
+    "- Tier 1 (Solo / Bronze): No subagents → you simulate specialization via persona instructions.",
+    "VERIFY THEN DELEGATE: Never assume subagent/tool availability. If uncertain, default to Tier 1.",
+    "SUBAGENT PRINCIPLES: Subagents are stateless cognitive specialists. Provide a complete work package. Prefer auditor model when possible.",
+    "PARALLELISM: Only run subagents in parallel when their work is independent (e.g., persona interviews, ideation perspectives, audits).",
+    "QUALITY GATES: If a deliverable is weak, record gaps in the doc and iterate. Do not pretend it is complete.",
+    "DT IDs: Maintain an append-only Idea Backlog with DT-### IDs tracked in the design doc Counters section."
+  ],
+  "functionDefinitions": [
+    {
+      "name": "ideaEntryTemplate",
+      "definition": "Use this template for each idea entry:\n- Idea ID: DT-###\n- Title: …\n- One-liner: …\n- Category: protocol | UX | authoring | observability | reliability | tooling | packaging | policy | other\n- Addresses: (pain/insight/constraint this targets)\n- Inspiration: (what prompted it)\n- Why it might work: …\n- Key risks: …\n- Notes: …\n\nID rule: Allocate the next DT-### from the design doc section `## Counters` (e.g., `Next DT ID: DT-001`) and increment it after each new idea."
+    },
+    {
+      "name": "clusterTemplate",
+      "definition": "Use this template for each cluster/theme:\n- Theme name: …\n- Why these ideas belong together (1 line): …\n- Member idea IDs: DT-…\n- What it enables: …\n- Biggest risk/unknown: …"
+    },
+    {
+      "name": "prototypeSpecTemplate",
+      "definition": "Use this template for a minimal prototype spec:\n## Goal\n## Non-goals\n## Prototype learning question\n## Prototype artifact (what you built)\n## Proposed API / Interface (if applicable)\n- Inputs\n- Outputs\n- Invariants\n- Error cases (as data)\n## Migration strategy\n## Pre-mortem (how this fails)\n## Falsification criteria\n## Test plan\n"
+    },
+    {
+      "name": "testPlanTemplate",
+      "definition": "Use this template for the test/dogfood plan:\n## Scenarios to run\n## Agents/models/IDEs to test\n## Pass/fail metrics\n## Instrumentation needed\n## Rollback/contingency\n"
+    },
+    {
+      "name": "personaInterviewTemplate",
+      "definition": "Use this template for each persona interview (real or simulated):\n## Persona\n- Name: …\n- Context: …\n- Goals: …\n- Pains: …\n- Constraints: …\n\n## Interview (Q → A)\n- Q: What does success look like?\n  A: …\n- Q: What would make you reject this direction?\n  A: …\n- Q: What is your biggest fear/risk?\n  A: …\n- Q: What tradeoff do you care about most?\n  A: …\n- Q: What would delight you?\n  A: …\n\n## Notes\n- …\n\nLabel: If simulated, state clearly: `Evidence Type: Simulated`."
+    }
+  ],
+  "steps": [
+    {
+      "id": "phase-0-triage",
+      "title": "Phase 0: Minimal Intake (Autonomous)",
+      "prompt": "Capture the minimal human input needed to run autonomously.\n\nDeliverable (bullets under headings):\n## Problem / Opportunity\n- …\n## Constraints\n- Hard: …\n- Soft: …\n## Non-negotiable principles\n- …\n## Anti-goals / out-of-scope\n- …\n## Intended deliverable\n- …\n\nAutonomy rule: If something is missing, write it as an explicit assumption with a validation plan (do not invent).\n\nEnd with 'Next Input' checklist (only items truly required to proceed).",
+      "agentRole": "You are a product/engineering facilitator. Keep intake minimal and make assumptions explicit.",
+      "requireConfirmation": true
+    },
+    {
+      "id": "phase-0a-init-design-doc",
+      "title": "Phase 0A: Initialize Canonical Design Doc (Resumption Anchor)",
+      "prompt": "Create a single canonical design doc that will be the source of truth for the rest of this workflow.\n\nDeliverable:\n1) Choose and confirm a file path: `designDocPath`\n   - Default suggestion: `ideas/<short-slug>-design-thinking.md`\n2) Create the file (if missing) and write the initial skeleton sections:\n   - Context / Problem\n   - Constraints / Principles / Anti-goals\n   - Execution Tier\n   - Assumption Register\n   - Persona Bank\n   - Empathy Findings (Observations / Insights)\n   - POV + HMW\n   - Success Criteria\n   - Idea Backlog (append-only)\n   - Clusters + Scoring\n   - Decision Log\n   - Pre-mortem + Falsification\n   - Prototype Spec\n   - Test Plan\n   - Iteration Notes\n   - Counters (DT IDs)\n3) Persist Phase 0 outputs into the doc under the appropriate sections.\n\nIn `Counters`, initialize:\n- Next DT ID: DT-001\n\nState (must be explicit):\n- Set `designDocPath` to the chosen path.\n\nChat output: only confirm `designDocPath` and list any missing inputs before proceeding.",
+      "agentRole": "You create a resumption-proof anchor artifact and ensure all future work is doc-first.",
+      "requireConfirmation": true
+    },
+    {
+      "id": "phase-0b-environment-handshake",
+      "title": "Phase 0B: Agent Cascade Boot Check (Tier 1/2/3)",
+      "prompt": "Determine which execution tier you can use (Agent Cascade Protocol).\n\nDoc-first: In `designDocPath`, fill `## Execution Tier` with:\n- Tier chosen: 1 (Solo) | 2 (Proxy) | 3 (Delegation)\n- Evidence: what you observed\n- Constraints: what you cannot do\n\nBoot check procedure:\n1) Check if subagents exist in your environment.\n   - If none: Tier 1.\n2) If subagents exist, ask one subagent to call `workflow_list`.\n   - If they can call WorkRail MCP tools: Tier 3.\n   - If they cannot: Tier 2.\n\nRule: If anything is uncertain, default to Tier 1.\n\nChat: one line stating Tier chosen.",
+      "agentRole": "You are verifying capabilities before using them. Prevent hallucinated delegation.",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-1a-plan-personas",
+      "title": "Phase 1A: Plan Empathy via Persona Simulation",
+      "prompt": "Plan how you will gather empathy signals with minimal human input.\n\nDoc-first: In `designDocPath`, update `## Persona Bank` with 3–5 personas you will interview (real or simulated).\nFor each persona: name, why it matters, what it can falsify.\n\nRequired personas (minimum):\n- Primary User\n- Power User\n- Skeptic (Stakeholder / Risk owner)\n- Operator (maintenance/support)\n\nEnd with a short plan: which 3 questions matter most for this problem.",
+      "agentRole": "You plan a structured empathy strategy. No ideation yet.",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-1b-persona-interviews",
+      "title": "Phase 1B: Conduct Persona Interviews (Tiered)",
+      "prompt": "Conduct persona interviews and capture results. Label simulated evidence clearly.\n\nTiered execution:\n\nIF Tier 3 (Delegation):\n- Spawn 4 persona interviewers SIMULTANEOUSLY (Primary User, Power User, Skeptic, Operator).\n- Provide each a complete work package including: problem, constraints, intended deliverable, and the personaInterviewTemplate.\n- Each must return a structured deliverable with the template and `Evidence Type: Simulated` unless you have real quotes.\n\nIF Tier 2 (Proxy):\n- Ask subagents to roleplay each persona via natural language.\n- You (main agent) transcribe results into the doc using personaInterviewTemplate and label as simulated.\n\nIF Tier 1 (Solo):\n- ACT AS each persona sequentially. Use personaInterviewTemplate. Label as simulated.\n\nDoc-first: Append interviews under `## Persona Bank` (or a `## Persona Interviews` section).\nChat: short recap + 'Next Input' checklist.",
+      "agentRole": "You are gathering empathy signals via structured interviews. Be explicit about what is simulated.",
+      "requireConfirmation": false,
+      "functionReferences": ["personaInterviewTemplate()"]
+    },
+    {
+      "id": "phase-1c-synthesize-empathy",
+      "title": "Phase 1C: Synthesize Empathy (Observations → Insights)",
+      "prompt": "Synthesize persona interviews into grounded observations and insights.\n\nDoc-first: Update `## Empathy Findings` with:\n- Observations (5–10)\n- Insights (5–10), each tied to at least one observation\n- Unknowns\n- Assumption Register updates (explicit)\n\nAdversarial/self-reflection (required):\n- What might we be misreading?\n- What evidence would contradict the top insight?\n- What could be \"working as designed\" (not actually a problem)?\n\nOptional Tier 3 audit (recommended for important decisions):\n- Spawn 2 auditors simultaneously (Completeness focus, Depth focus) to audit your empathy synthesis and call out gaps/assumptions.\n- If Tier 2/1: self-audit with the same questions.\n\nChat: short recap + 'Next Input' checklist.",
+      "agentRole": "You converge empathy into actionable insights and explicit assumptions.",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-2-define",
+      "title": "Phase 2: Define (POV + HMW + Success Criteria)",
+      "prompt": "Turn empathy into a precise definition.\n\nDoc-first: Update `## POV + HMW` and `## Success Criteria` with:\n## POV\n- [Persona] needs [need] because [insight].\n\n## Problem statement (2–4 lines)\n- …\n\n## Alternative framings (2)\n- If we are wrong about the core problem, what else could it be?\n- What would be a radically simpler interpretation?\n\nAnchoring check: If you cannot produce two plausible alternative framings, say so explicitly and go back to Empathy Findings to improve them.\n\n## How might we… (3–7)\n- …\n\n## Success criteria\n- …\n\n## Key tensions / tradeoffs\n- …\n\n## Riskiest assumption\n- …\n\n## What would change our mind?\n- …\n\nChat: short recap + 'Next Input' checklist.",
+      "agentRole": "You create a crisp, testable framing and protect against anchoring.",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-2b-challenge-framing",
+      "title": "Phase 2B: Adversarial Challenge the Framing (Tiered)",
+      "prompt": "Challenge your own framing before ideation.\n\nIF Tier 3 (Delegation):\n- Delegate to the **WorkRail Executor** to run `routine-hypothesis-challenge` at rigor=3.\n- Mission: Attack POV, HMW, riskiest assumption, and success criteria.\n- Provide complete context: paste the relevant doc sections.\n- Incorporate valid critique into the doc.\n\nIF Tier 2 (Proxy):\n- Ask a subagent to act as an adversarial critic (no WorkRail tools).\n- You synthesize their critique and update the doc.\n\nIF Tier 1 (Solo):\n- ACT AS a skeptical reviewer. Try to prove the framing is wrong.\n\nDoc-first: append a short `## Framing Challenge` section with findings and changes made.",
+      "agentRole": "You are preventing premature convergence by adversarially testing the framing.",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-3-ideation-loop",
+      "type": "loop",
+      "title": "Phase 3: Ideation (4 Rounds)",
+      "loop": { "type": "for", "count": 4, "maxIterations": 4, "iterationVar": "ideationRound" },
+      "body": [
+        {
+          "id": "ideation-round-1",
+          "title": "Ideation Round 1: General Divergence",
+          "runCondition": { "var": "ideationRound", "equals": 1 },
+          "prompt": "Generate a broad set of ideas directly relevant to the defined problem.\n\nDoc-first: Append ideas to Idea Backlog in `designDocPath` using ideaEntryTemplate.\nDT IDs: allocate next DT-### from `## Counters`.\n\nTiered execution:\n- Tier 3: Spawn 2–3 ideators SIMULTANEOUSLY with different perspectives (Simplicity, Maintainability, Pragmatic). Synthesize into one backlog.\n- Tier 2: Ask subagents to brainstorm per perspective; you curate into the backlog.\n- Tier 1: ACT AS each perspective sequentially; then curate.\n\nReflection (brief, required):\n- Which category is underrepresented?\n- What is the simplest idea we are dismissing too quickly?\n- What assumption seems to be driving most ideas?",
+          "agentRole": "Divergent ideation with explicit perspective diversity.",
+          "functionReferences": ["ideaEntryTemplate()"]
+        },
+        {
+          "id": "ideation-round-2",
+          "title": "Ideation Round 2: Orthogonal / Unrelated Inspiration",
+          "runCondition": { "var": "ideationRound", "equals": 2 },
+          "prompt": "Generate ideas that are intentionally NOT obvious extensions of Round 1.\n\nDoc-first: Append ideas with DT IDs.\nTechniques: analogies, inversions, constraint creativity.\n\nConstraint inversion (required): add 5 ideas that assume a hard constraint and explain why it changes the solution.\n\nTiered execution:\n- Tier 3: Spawn 2 ideators SIMULTANEOUSLY (Contrarian, Analogist). Synthesize.\n- Tier 2: Proxy brainstorm via subagents; you curate.\n- Tier 1: Do it yourself, explicitly switching personas.\n\nReflection (brief, required):\n- Which analogy created the most leverage?\n- What is the simplest idea we are dismissing too quickly?\n- What assumption seems to be driving most ideas?",
+          "agentRole": "Break local maxima with orthogonal ideation.",
+          "functionReferences": ["ideaEntryTemplate()"]
+        },
+        {
+          "id": "ideation-round-3",
+          "title": "Ideation Round 3: Combine / Mutate / Package",
+          "runCondition": { "var": "ideationRound", "equals": 3 },
+          "prompt": "Generate ideas that combine and strengthen the existing backlog.\n\nDoc-first: Append derived ideas with DT IDs.\nAlso produce 5 concept packages referencing member DT-IDs.\n\nTiered execution:\n- Tier 3: Spawn 2 ideators SIMULTANEOUSLY (Systems, Pragmatic). Synthesize.\n- Tier 2: Proxy brainstorm; you curate.\n- Tier 1: Do it yourself.\n\nReflection (brief, required):\n- What critical primitive is missing?\n- What is the simplest idea we are dismissing too quickly?\n- What assumption seems to be driving most ideas?",
+          "agentRole": "Converge divergent ideas into packages without prematurely selecting.",
+          "functionReferences": ["ideaEntryTemplate()"]
+        },
+        {
+          "id": "ideation-round-4",
+          "title": "Ideation Round 4: Coverage Sweep & Blind Spots",
+          "runCondition": { "var": "ideationRound", "equals": 4 },
+          "prompt": "Audit the idea space for blind spots and fill gaps.\n\nChecklist: protocol, resumption, authoring UX, validation, dashboard/observability, model variability, external workflow packaging, loops correctness, capability negotiation.\n\nDoc-first: Add 10–15 missing ideas with DT IDs.\nAdd Coverage Map: Dimension | coverage (low/med/high) | top DT-IDs.\n\nTiered execution:\n- Tier 3: Optional parallel audit: one ideator does blind spots; another acts as skeptic.\n- Tier 2: Proxy.\n- Tier 1: Solo.\n\nReflection (brief, required):\n- Where would we regret not exploring further?\n- What category did we avoid because it felt uncomfortable or \"too big\"?\n- What assumption seems to be driving most ideas?",
+          "agentRole": "Ensure coverage and surface blind spots.",
+          "functionReferences": ["ideaEntryTemplate()"]
+        }
+      ]
+    },
+    {
+      "id": "phase-4-synthesize",
+      "title": "Phase 4: Synthesize (Cluster + Prioritize)",
+      "prompt": "Synthesize the backlog into themes and a shortlist.\n\nDoc-first: Update `## Clusters + Scoring` with:\n- 5–10 clusters (use clusterTemplate)\n- Candidate directions (top 3–7) with North Star + scoring (Impact/Confidence/Migration/Model-robustness/Time-to-value)\n- Shortlist (2–3)\n\nAdversarial challenge (brief):\n- Argue the top direction is wrong. What would a skeptic say?\n- What is the strongest alternative and why might it win?\n\nTiered optional validation:\n- Tier 3: Spawn **two WorkRail Executor delegations** in parallel:\n  - WorkRail Executor #1: run `routine-plan-analysis` to audit shortlist completeness, scoring, and migration risk.\n  - WorkRail Executor #2: run `routine-hypothesis-challenge` at rigor=3 to adversarially challenge the top direction.\n- Tier 2/1: self-audit using the same prompts.\n\nEnd with a short Decision Log entry: what you chose and why.",
+      "agentRole": "You converge and make tradeoffs explicit.",
+      "requireConfirmation": false,
+      "functionReferences": ["clusterTemplate()"]
+    },
+    {
+      "id": "phase-4a-quality-gate",
+      "title": "Phase 4A: Synthesis Quality Gate",
+      "prompt": "Verify readiness to prototype.\n\nDoc-first: Add `## Synthesis Quality Gate` checklist with ✅/❌:\n- POV exists\n- 3–7 HMW exist\n- Success criteria exist\n- Key tensions exist\n- Idea Backlog has breadth\n- Shortlist (2–3) exists\n- Top direction has at least one falsifiable learning question\n\nIf FAIL: list what is missing and fix it before proceeding.",
+      "agentRole": "You enforce blueprint quality.",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-4b-premortem-falsification",
+      "title": "Phase 4B: Pre-mortem & Falsification",
+      "prompt": "Before prototyping, add a pre-mortem and explicit falsification criteria for the top direction. Treat directions as hypotheses; tests exist to falsify.\n\nDoc-first: Update `## Pre-mortem + Falsification` with:\n- Pre-mortem (top 5): failure mode | why | mitigation\n- Falsification criteria (1–3): if X happens, we pivot/stop because…\n\nReflection (brief):\n- What second-order effect is most dangerous?\n- What would we regret not testing?",
+      "agentRole": "You pressure-test before investing in a prototype.",
+      "requireConfirmation": false
+    },
+    {
+      "id": "phase-5-prototype",
+      "title": "Phase 5: Prototype (Learning Artifact)",
+      "prompt": "Build a low-fidelity prototype intended to answer one learning question.\n\nChoose ONE prototype artifact:\n- Storyboard (6 panels)\n- Concierge script (manual end-to-end flow)\n- API sketch (request/response examples + mocked outputs)\n- Wireframe (markdown screens + interactions)\n\nDoc-first: Write the full prototype spec and artifact into `designDocPath` using prototypeSpecTemplate.\nInclude falsification criteria (copy verbatim, then refine).\n\nReflection (brief):\n- What is the smallest slice that still proves/disproves the concept?\n- If falsified, what is the next-best direction from the shortlist?",
+      "agentRole": "You create a falsifiable learning prototype.",
+      "requireConfirmation": false,
+      "functionReferences": ["prototypeSpecTemplate()"]
+    },
+    {
+      "id": "phase-6-test-plan",
+      "title": "Phase 6: Test Plan (Dogfood + Validation)",
+      "prompt": "Define how to test the prototype with real usage constraints.\n\nDoc-first: Write the test plan into `designDocPath` using testPlanTemplate.\nMust include: Claude + one weaker model (GPT/Grok/Gemini) in the model/IDE matrix.\n\nReflection (brief):\n- If the test fails, what do we do next (iterate, pivot, stop)?",
+      "agentRole": "You design an evidence-producing test plan.",
+      "requireConfirmation": false,
+      "functionReferences": ["testPlanTemplate()"]
+    },
+    {
+      "id": "phase-7-iterate-loop",
+      "type": "loop",
+      "title": "Phase 7: Iterate (Test → Learn → Refine)",
+      "loop": { "type": "for", "count": 2, "maxIterations": 2, "iterationVar": "iteration" },
+      "body": [
+        {
+          "id": "iteration-test",
+          "title": "Iteration {{iteration}}: Dogfood Test & Capture Feedback",
+          "prompt": "Run a lightweight dogfood test of the prototype (thought experiment, adversarial critique, stakeholder Q&A simulation).\n\nDoc-first: Append to `## Iteration Notes`:\n- What we tried\n- What worked\n- What failed\n- Confusions/friction\n- New insights\n\nChat: short recap only.",
+          "agentRole": "You generate learning signals cheaply."
+        },
+        {
+          "id": "iteration-learn-update",
+          "title": "Iteration {{iteration}}: Learn & Update Artifacts",
+          "prompt": "Based on feedback, update the key artifacts in `designDocPath` and add `## Iteration {{iteration}}: Updates`.\n\nDeliverable:\n- Changes made (POV/HMW/shortlist/prototype/test plan)\n- Rationale\n\nReflection (brief):\n- What surprised us?\n- What did we previously believe that is now false?",
+          "agentRole": "You update artifacts based on learning."
+        }
+      ]
+    },
+    {
+      "id": "phase-8-handoff",
+      "title": "Phase 8: Final Summary & Handoff",
+      "prompt": "Finalize the design doc and produce a crisp handoff.\n\nDoc-first: Add `## Final Summary` with:\n- Problem statement\n- POV/HMW\n- Chosen direction + why\n- Falsification criteria\n- Prototype artifact summary\n- Test plan summary\n- What we learned across iterations\n- Next actions\n\nChat: one-paragraph summary + path to `designDocPath`.",
+      "agentRole": "You conclude the autonomous design thinking cycle with a handoff.",
+      "requireConfirmation": false
+    }
+  ]
+}