@blxzer/cursor-trellis 0.1.0

package/dist/templates/trellis/workflow.md

@@ -0,0 +1,960 @@
+# Development Workflow
+
+---
+
+## Core Principles
+
+1. **Plan before code** — figure out what to do before you start
+2. **Specs injected, not remembered** — guidelines are injected via hook/skill, not recalled from memory
+3. **Persist everything** — research, decisions, and lessons all go to files; conversations get compacted, files don't
+4. **Incremental development** — one task at a time
+5. **Capture learnings** — after each task, review and write new knowledge back to spec
+
+---
+
+## Trellis System
+
+### Developer Identity
+
+On first use, initialize your identity:
+
+```bash
+python3 ./.trellis/scripts/init_developer.py <your-name>
+```
+
+Creates `.trellis/.developer` (gitignored) + `.trellis/workspace/<your-name>/`.
+
+### Spec System
+
+`.trellis/spec/` holds coding guidelines organized by package and layer.
+
+- `.trellis/spec/<package>/<layer>/index.md` — entry point with **Pre-Development Checklist** + **Quality Check**. Actual guidelines live in the `.md` files it points to.
+- `.trellis/spec/guides/index.md` — cross-package thinking guides.
+
+```bash
+python3 ./.trellis/scripts/get_context.py --mode packages   # list packages / layers
+```
+
+**When to update spec**: new pattern/convention found · bug-fix prevention to codify · new technical decision.
+
+### Task System
+
+Every task has its own directory under `.trellis/tasks/{MM-DD-name}/` holding `task.json`, `prd.md`, optional `design.md`, optional `implement.md`, optional `research/`, and context manifests (`implement.jsonl`, `check.jsonl`) for sub-agent-capable platforms.
+
+Optional copied templates live under `.trellis/tasks/templates/` (for example `release-readiness/` and `release-execution/`). `task.py create` does not auto-apply them; copy files into a new task when starting that workflow.
+
+```bash
+# Task lifecycle
+python3 ./.trellis/scripts/task.py create "<title>" [--slug <name>] [--parent <dir>]
+python3 ./.trellis/scripts/task.py dashboard             # show Task Dashboard without mutating state
+python3 ./.trellis/scripts/task.py select <name>         # select task for this live session
+python3 ./.trellis/scripts/task.py selected --source     # show selected task and source
+python3 ./.trellis/scripts/task.py start-execution <name> --check
+python3 ./.trellis/scripts/task.py start-execution <name> --approved
+python3 ./.trellis/scripts/task.py exit                  # clear selected task without changing status
+python3 ./.trellis/scripts/task.py archive <name>        # move to archive/{year-month}/
+python3 ./.trellis/scripts/task.py list [--mine] [--status <s>]
+python3 ./.trellis/scripts/task.py list-archive
+python3 ./.trellis/scripts/task.py add-subtask <parent> <child>
+python3 ./.trellis/scripts/task.py set-child-state <parent> <child> review --evidence verify.md
+python3 ./.trellis/scripts/task.py prepare-child-worktree <parent> <child> --branch <child-branch>
+python3 ./.trellis/scripts/task.py integrate-child <parent> <child> accepted --evidence handoff.md --ref <child-ref>
+
+# Code-spec context (injected into implement/check agents via JSONL).
+# `implement.jsonl` / `check.jsonl` are seeded on `task create` for sub-agent-capable
+# platforms; the AI curates real spec + research entries during planning when needed.
+python3 ./.trellis/scripts/task.py add-context <name> <action> <file> <reason>
+python3 ./.trellis/scripts/task.py list-context <name> [action]
+python3 ./.trellis/scripts/task.py validate <name>
+
+# Task metadata
+python3 ./.trellis/scripts/task.py set-branch <name> <branch>
+python3 ./.trellis/scripts/task.py set-base-branch <name> <branch>    # PR target
+python3 ./.trellis/scripts/task.py set-scope <name> <scope>
+
+# Hierarchy (parent/child)
+python3 ./.trellis/scripts/task.py add-subtask <parent> <child>
+python3 ./.trellis/scripts/task.py remove-subtask <parent> <child>
+
+# PR creation
+python3 ./.trellis/scripts/task.py create-pr [name] [--dry-run]
+```
+
+> Run `python3 ./.trellis/scripts/task.py --help` to see the authoritative, up-to-date list.
+
+**Selected-task mechanism**: entering a Trellis project activates framework context, but every new live session starts with `Selected task: none`. `task.py create` creates artifacts only. `task.py select <task>` writes a per-session `selected_task` pointer without changing `task.json.status`. `task.py selected --source` reports that pointer. `task.py exit` clears it without changing status. `task.py start-execution <task> --check` verifies execution readiness without mutation; `task.py start-execution <task> --approved` is the explicit execution boundary and may flip `planning` to `in_progress`. `task.py archive <task>` writes `status=completed`, moves the directory to `archive/`, and deletes runtime session files that still point at the archived task.
+
+### Workspace System
+
+Records every AI session for cross-session tracking under `.trellis/workspace/<developer>/`.
+
+- `journal-N.md` — session log. **Max 2000 lines per file**; a new `journal-(N+1).md` is auto-created when exceeded.
+- `index.md` — personal index (total sessions, last active).
+
+```bash
+python3 ./.trellis/scripts/add_session.py --title "Title" --commit "hash" --summary "Summary"
+```
+
+### Context Script
+
+```bash
+python3 ./.trellis/scripts/get_context.py                            # full session runtime
+python3 ./.trellis/scripts/get_context.py --mode packages            # available packages + spec layers
+python3 ./.trellis/scripts/get_context.py --mode phase --step <X.Y>  # detailed guide for a workflow step
+python3 ./.trellis/scripts/get_context.py --mode retrieval-pack --json --input <evidence.json>  # score collected evidence (not default --json)
+python3 ./.trellis/scripts/get_context.py --mode retrieval-pack --platform cursor --json  # platform-adaptive routing
+```
+
+**Platform-adaptive routing:** pass `--platform <name>` (cursor, claude-code, codex) when calling `--mode retrieval-pack`. This enables platform-specific tool recommendations (e.g., Cursor uses `@codebase` instead of `fast-context`). Hooks auto-detect the platform; manual calls should include it when the agent knows the host platform.
+
+**Evidence scoring:** default `--json` returns `retrievalGuide` only. After collecting artifact search, session memory, smart-search manifests under `{TASK}/research/smart-search/`, or codebase candidates, run **`--mode retrieval-pack`** with `--input` or stdin JSON. See `research/evidence-scoring-integration.md` in the active task or archived `06-15-child-phase2-evidence-scoring`.
+
+**Research-end hook (Cursor `stop` / Claude `Stop`):** when the selected task has `{TASK}/research/*.md` or `research/smart-search/`, `.cursor/hooks/research-end-retrieval-pack.py` (or `.claude/hooks/…`) may write `{TASK}/research/retrieval-pack-latest.json` via `get_context --mode retrieval-pack --platform <auto-detected>`. Default session JSON is unchanged; use the file when closing research or before Phase 3.1.
+
+**Retrieval daily guide:** `.trellis/spec/guides/retrieval-daily-guide.md` — when to use rg, codegraph, fast-context-mcp, smart-search-cli (and Cursor web fallback), artifact/session memory, codebase router (suggest-only), and explicit retrieval-pack scoring.
+
+**Cursor subagent dispatch:** `.trellis/spec/guides/cursor-subagent-policy.md` — `trellis-research` / `trellis-implement` / `trellis-check`; Parent child default **Task** `trellis-implement` from Parent session (`generate-child-prompt --mode subagent`). **Cursor++ BYOK:** per-type models via `.trellis/local/cursor2plus/` + user/project JSON maps (not committed slugs). **Native Cursor API:** frontmatter `model:` on agents still works. PRD Grill stays in `trellis-brainstorm`, not a subagent. **Cursor++:** compatible v0.0.11+ (SubAgent readonly bug fixed).
+
+---
+
+<!--
+  WORKFLOW-STATE BREADCRUMB CONTRACT (read this before editing the tag blocks below)
+
+  The [workflow-state:STATUS] blocks embedded in the ## Phase Index section
+  below are the SINGLE source of truth for the per-turn `<workflow-state>`
+  breadcrumb that every supported AI platform's UserPromptSubmit hook
+  reads. inject-workflow-state.py (Python platforms) and
+  inject-workflow-state.js (OpenCode plugin) only parse them — there is no
+  fallback dict baked into the scripts after v0.5.0-rc.0.
+
+  STATUS charset: [A-Za-z0-9_-]+. When the hook can't find a tag, it
+  degrades to a generic "Refer to workflow.md for current step." line —
+  intentionally visible so users notice and fix a broken workflow.md.
+
+  INVARIANT (test/regression.test.ts):
+    Every workflow-walkthrough step marked `[required · once]` must have a
+    matching enforcement line in its phase's [workflow-state:*] block. The
+    breadcrumb is the only per-turn channel; if a mandatory step isn't
+    mentioned there, the AI silently skips it (Phase 1 planning gate
+    skip and Phase 3.4 commit skip both manifested via this gap).
+
+  TAG ↔ PHASE scoping:
+    [workflow-state:no_task]      → framework active, selected task none; before Phase 1
+    [workflow-state:planning]     → all of Phase 1 (status='planning')
+    [workflow-state:planning-inline] → Codex inline variant of Phase 1
+    [workflow-state:in_progress]  → Phase 2 + Phase 3.1-3.4
+                                    (status stays 'in_progress' from
+                                    start-execution --approved until task.py archive)
+    [workflow-state:in_progress-inline] → Codex inline variant of Phase 2/3
+    [workflow-state:completed]    → currently DEAD: cmd_archive flips
+                                    status and moves the dir in the same
+                                    call, so the resolver loses the
+                                    pointer (block kept for a future
+                                    explicit in_progress→completed
+                                    transition)
+
+  Editing checklist:
+    - When you change a [workflow-state:STATUS] block, also check the
+      matching phase's `[required · once]` walkthrough steps for sync
+    - Run `trellis update` after editing to push the new bodies to
+      downstream user projects (block-level managed replacement)
+    - Full runtime contract:
+      .trellis/spec/cli/backend/workflow-state-contract.md
+-->
+
+## Phase Index
+
+```
+Phase 1: Plan    → classify, get task-creation consent, then write planning artifacts
+Phase 2: Execute → implement only after task status is in_progress
+Phase 3: Finish  → verify, record learning decision, commit, and guarded archive
+```
+
+### Request Triage (mandatory before any work)
+
+**Every turn that could produce work must be classified before acting.** This is a hard gate, not a suggestion. Resolve the request against the Task Ladder decision tree below, then emit the classification mark. If you cannot classify, you have not understood the request — ask a clarifying question instead of starting work.
+
+Decision tree (first match wins):
+
+1. **No durable project change** (conversation, status, explanation, read-only lookup, tiny one-turn action) → `No Task`.
+2. **Underspecified small request, no task yet** (needs focused clarification or decision pressure before work exists) → `Micro-Grill` (load `trellis-micro-grill`).
+3. **Low-risk durable work, narrow scope, local validation, no shared contract** → `Lite Task`.
+4. **Durable code/template/runtime/workflow/cross-file behavior, or framework semantics** → `Full Task`.
+5. **Multiple independent deliverables, staged/parallel execution, or final integration authority** → `Parent Task / Child Tasks`.
+
+Classification mark (R2 — visible audit trail). Start your reply with one line in this exact shape:
+
+```
+[Triage: <Mode>] <one-sentence reason citing the trigger signal>
+```
+
+- `<Mode>` ∈ `No Task | Micro-Grill | Lite | Full | Parent`.
+- The reason must reference the trigger signal from the Task Ladder table (e.g. "cross-file workflow change", "read-only explanation"). Vague reasons like "looks complex" are not acceptable.
+- For `No Task` turns the mark is still required — it is how the user audits that you actually classified rather than skipped.
+
+Consent gate. After classifying into any mode that creates a task, ask the user for task-creation consent before creating any Trellis artifact. User approval to create a task is **not** approval to start implementation — planning still happens first. If the user declines a task for a simple request, skip Trellis for this session.
+
+Selected-task continuity. When a `selected_task` already exists, do not rerun global classification on every follow-up; continue inside the selected task unless a strong conflict exists (explicit exit/switch/create language, out-of-scope request, different artifact/archive target, new independent deliverable, contract-changing request, or evidence pollution risk).
+
+### Task Ladder And Routing
+
+Classify by risk and persistence, not raw effort size. A short change to durable framework semantics can require a Full Task; a long conversation can remain No Task when it leaves no durable project state.
+
+| Mode | Use when | Trigger signals | Durable artifacts |
+| --- | --- | --- | --- |
+| No Task | Conversation, status, explanation, read-only lookup, or a tiny one-turn action with no durable project change. | explain / status / lookup / read-only / one-liner | None. No archive unless upgraded. |
+| Micro-Grill | The user needs focused clarification, decision pressure, or a small requirement interrogation before deciding whether work exists. | small + underspecified / "depends" / needs clarification / decision tree first | Usually none. Upgrade before durable edits, validation, gates, or archive evidence. |
+| Lite Task | Low-risk durable work with narrow scope, local validation, and no shared contract change. | low-risk / single file / local validation / no contract / narrow scope | `task.json`, `prd.md`, `verify.md`, and archive evidence. |
+| Full Task | Durable code, template, runtime, workflow, or cross-file behavior where design, execution strategy, validation, or reviewer gates matter. | cross-file / framework semantics / contract change / template / runtime / workflow / multi-file behavior | `prd.md`, `design.md`, `implement.md`, `verify.md`, Development Strategy Contract, `verification_profile`, `quality_gates`, and archive evidence. |
+| Parent Task / Child Tasks | One request contains independent deliverables, staged execution, parallel execution, or final integration authority that must be owned by a Parent. | multiple independent deliverables / staged / parallel / integration authority | Parent `task-map.md`, Child task artifacts, Child handoff evidence, Parent final integration evidence. |
+
+Default Trellis framework semantics, task model, platform adapters, MCP/capability setup, runtime integration, retrieval/graph tooling, Parent/Child orchestration, and quality-gate work to Full Task or higher.
+
+When `Selected task: none`, enter repo-first routing: read local instructions/workflow evidence, run `task.py dashboard` when useful, classify the request on the ladder, and ask for task-creation consent before creating artifacts. Do not auto-select an existing task.
+
+When a `selected_task` already exists, do not rerun global classification on every follow-up. Continue inside the selected task unless a strong conflict exists: explicit exit/switch/create language, an out-of-scope request, a different task artifact or archive target, a new independent deliverable, a contract-changing request, or evidence pollution risk. A contract-changing request under `selected_task` routes to that selected task's Planning flow unless the user explicitly switches or creates another task.
+
+### Upgrade / Downgrade Rules
+
+- No Task -> Micro-Grill when the turn needs structured clarification or a decision tree before work can be safely classified.
+- Micro-Grill -> Lite/Full when the outcome needs persistent task artifacts, repo edits, validation evidence, quality gates, or archive.
+- Lite -> Full when scope touches shared contracts, multi-file behavior, framework semantics, platform/runtime/capability assumptions, `verification_profile`, `quality_gates`, or rollback-sensitive validation.
+- Full -> Parent/Child only when the work has independent deliverables, staged execution, parallel execution, or Parent-controlled final integration needs.
+
+Before executing an upgrade that creates artifacts, changes task mode, adds gates, changes `verification_profile` or capabilities, or changes approval requirements, get explicit user confirmation. Every downgrade needs explicit user confirmation because it reduces artifact, gate, validation, or approval rigor.
+
+### Task Ladder quick routing
+
+| Situation | Action |
+|-----------|--------|
+| No selected task + small unclear ask | `trellis-micro-grill` |
+| No selected task + need dashboard | `trellis-start` |
+| Selected task + resume step | `trellis-continue` |
+| Planning / PRD | `trellis-brainstorm` |
+| Parent with parallel children | `generate-child-prompt --mode subagent`; writable Agent; see `.trellis/spec/guides/cursor-subagent-policy.md` |
+
+Details: archived `06-15-child-phase3-task-ladder` → `research/task-ladder-iteration.md`.
+
+### Planning Artifacts
+
+- `prd.md` — requirements, constraints, and acceptance criteria. Do not put technical design or execution checklists here.
+- `design.md` — technical design for complex tasks: boundaries, contracts, data flow, tradeoffs, compatibility, rollout / rollback shape.
+- `implement.md` — execution plan for complex tasks: ordered checklist, Development Strategy Contract, validation commands, review gates, and rollback points.
+- `implement.jsonl` / `check.jsonl` — spec and research manifests for sub-agent context. They do not replace `implement.md`.
+- `verification_profile` / `quality_gates` — gate policy belongs in task artifacts and `task.json`; `task.json.quality_gate_results` is compact machine-checkable state, not human review prose.
+- Lightweight tasks may be PRD-only. Complex tasks must have `prd.md`, `design.md`, and `implement.md` before `task.py start-execution --check`.
+- `start-execution` planning gates (`requirements-review`, `architecture-review` when enabled) auto-record on `--approved` when artifacts pass CLI checks; use `record-gate` only as a manual override.
+
+### Parent / Child Task Trees
+
+Use a parent task when one user request contains several independently verifiable deliverables. The parent task owns the source requirement set, the task map, cross-child acceptance criteria, and final integration review; it normally should not be the implementation target unless it also has direct work.
+
+Use child tasks for deliverables that can be planned, implemented, checked, and archived independently. Parent/child structure is not a dependency system: if one child must wait for another, write that ordering in the child `prd.md` / `implement.md` and keep each child's acceptance criteria testable.
+
+Create new children with `task.py create "<title>" --slug <name> --parent <parent-dir>`. Link existing tasks with `task.py add-subtask <parent> <child>`, and unlink mistakes with `task.py remove-subtask <parent> <child>`.
+
+Child Workers report only Child-owned progress states with `task.py set-child-state`: `open`, `working`, `blocked`, or `review`. Parent-controlled setup and decisions use `task.py prepare-child-worktree` and `task.py integrate-child`: `changes`, `accepted`, `integrating`, `integrated`, or `cancelled`. Parent integration requires reviewed evidence, Child `verify.md` / `handoff.md`, and a short `--ref` for accepted/integrating/integrated states. Default Parent `merge_limit: 1` blocks more than one Child from being `integrating` at the same time.
+
+Integration is Parent/Child-only. Ordinary Lite and Full Tasks skip Integration and go from Verification / Review to Archive / Learning checks. A Child can provide evidence and request review, but cannot mark itself `changes`, `accepted`, `integrating`, `integrated`, or `cancelled`; only the Parent has integration authority. Parent integration is serial Git-ref integration by default: Child worktrees are prepared explicitly, Child decisions carry refs, `integrate-child ... integrated --execute-merge` runs an explicit no-commit merge when requested, and every decision respects `merge_limit: 1` and writes conflicts, merge decisions, and acceptance rationale to `task-map.md` Event Log.
+
+### Parent reviewer orchestration (inline or optional subagent)
+
+Parent sessions can productize child dispatch and review without a new agent runtime:
+
+```bash
+python3 ./.trellis/scripts/task.py parent-status <parent-task>
+python3 ./.trellis/scripts/task.py generate-child-prompt <parent-task> <child-task> --mode inline
+python3 ./.trellis/scripts/task.py review-child <parent-task> <child-task> --check --decision accept --ref <child-ref>
+python3 ./.trellis/scripts/task.py review-child <parent-task> <child-task> --decision accept --ref <child-ref>
+python3 ./.trellis/scripts/task.py review-child <parent-task> <child-task> --decision integrate-through --ref <child-ref>
+```
+
+- `generate-child-prompt` reads parent `task-map.md` for `depends_on` and `touches`, child artifacts, and optional parent `child-prompts.md`. Use `--mode subagent` only as a delivery hint when the platform can spawn subagents; inline mode remains the portable default.
+- `review-child` summarizes child `verify.md` / `handoff.md`, appends notes to parent `verify.md`, and can advance `accepted` / `integrating` / `integrated` in one flow (`--decision integrate-through`) while still using the same Stage 0 integration guards as `integrate-child`.
+- Reviewer quality gates (`child-review`, `parent-accepted`, `parent-integrated`) are **not** auto-recorded; the command prints optional `record-gate` hints only.
+
+<!-- Per-turn breadcrumb: shown when no task is selected (before Phase 1) -->
+
+[workflow-state:no_task]
+Trellis framework active. Selected task: none. Use `task.py dashboard` for routing; do not auto-select an existing task.
+MANDATORY TRIAGE (hard gate, not optional): every work-capable turn must be classified FIRST, before any action, into No Task / Micro-Grill / Lite Task / Full Task / Parent Task — see "Request Triage" in workflow.md for the decision tree. Emit the classification as the first line of your reply: `[Triage: <Mode>] <one-sentence reason citing the trigger signal>`. If you cannot classify, you have not understood the request — ask a clarifying question instead of starting work.
+After classifying into any mode that creates a task, ask the user for task-creation consent before creating any Trellis artifact. User consent to create a task is NOT consent to start implementation — planning still happens first.
+Underspecified small request with no task: load `trellis-micro-grill` before creating artifacts or upgrading the ladder.
+Framework refresh with no selected task: load `trellis-start` once; after the user selects a task, use `trellis-continue` for step-level resume—not `trellis-start`.
+[/workflow-state:no_task]
+
+### Phase 1: Plan
+- 1.0 Create task `[required · once]` (only after task-creation consent)
+- 1.1 Requirement exploration `[required · repeatable]` (`prd.md`; complex tasks also need `design.md` + `implement.md`)
+- 1.2 Research `[optional · repeatable]`
+- 1.3 Configure context `[conditional · once]` — Claude Code, Cursor, OpenCode, Codex, Kiro, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi
+- 1.4 Execution gate `[required · once]` (`task.py start-execution <task> --check`, explicit approval, then `--approved`; status → in_progress)
+- 1.5 Completion criteria
+
+<!-- Per-turn breadcrumb: shown throughout Phase 1 (status='planning') -->
+
+[workflow-state:planning]
+Load `trellis-brainstorm`; stay in planning.
+Lightweight: `prd.md` can be enough. Complex: finish `prd.md`, `design.md`, and `implement.md`; run `task.py start-execution <task> --check`, report PASS with task plus current contract/fingerprint context, and ask for explicit execution approval before `--approved`.
+Multi-deliverable scope: consider a parent task plus independently verifiable child tasks; dependencies must be written in child artifacts, not implied by tree position.
+Sub-agent mode: curate `implement.jsonl` and `check.jsonl` as spec/research manifests before start.
+[/workflow-state:planning]
+
+<!-- Per-turn breadcrumb: shown throughout Phase 1 when codex.dispatch_mode=inline.
+     Codex-only opt-in alternate to [workflow-state:planning]. The main agent
+     edits code directly in Phase 2, so jsonl curation is skipped —
+     the inline workflow loads `trellis-before-dev` instead of injecting JSONL
+     into a sub-agent. -->
+
+[workflow-state:planning-inline]
+Load `trellis-brainstorm`; stay in planning.
+Lightweight: `prd.md` can be enough. Complex: finish `prd.md`, `design.md`, and `implement.md`; run `task.py start-execution <task> --check`, report PASS with task plus current contract/fingerprint context, and ask for explicit execution approval before `--approved`.
+Multi-deliverable scope: consider a parent task plus independently verifiable child tasks; dependencies must be written in child artifacts, not implied by tree position.
+Inline mode: skip jsonl curation; Phase 2 reads artifacts/specs via `trellis-before-dev`.
+[/workflow-state:planning-inline]
+
+### Phase 2: Execute
+- 2.1 Implement `[required · repeatable]`
+- 2.2 Quality check `[required · repeatable]`
+- 2.3 Rollback `[on demand]`
+
+<!-- Per-turn breadcrumb: shown while status='in_progress'.
+     Scope: all of Phase 2 + Phase 3.1-3.4 (status stays 'in_progress' from
+     start-execution --approved until task.py archive; only archive flips it). The body
+     therefore must cover every required step from implementation through
+                                    commit, including Phase 3.3 learning decision and Phase 3.4 commit. -->
+
+Sub-agent dispatch protocol applies to all platforms and all sub-agents, including class-2 Codex/Copilot/Gemini/Qoder and `trellis-research`: every dispatch prompt starts with `Selected task: <task path from task.py selected>` before role-specific instructions.
+
+[workflow-state:in_progress]
+Tools: `trellis-implement` / `trellis-research` are sub-agent types only (Task/Agent tool, NOT Skill; there is no skill by these names). `trellis-update-spec` is a skill for durable learning only. `trellis-check` exists as both; prefer the Agent form when verifying after code changes.
+Flow: `trellis-implement` -> `trellis-check` -> validation/evidence in `verify.md` -> learning decision -> commit (Phase 3.4) -> `task.py archive <task> --check` -> `/trellis:finish-work`.
+Execution boundary: implement only inside approved `prd.md`, `design.md`, `implement.md`, and Development Strategy Contract; stop and Return-to-Planning for scope, contract, gate, capability/runtime, Parent `contract_epoch`, Child boundary, selected-task fit, or non-implementation reviewer-gate changes.
+Main-session default: dispatch implement/check sub-agents. Sub-agent self-exemption: if already running as `trellis-implement`, do NOT spawn another `trellis-implement` or `trellis-check`; if already running as `trellis-check`, do NOT spawn another `trellis-check` or `trellis-implement`. Dispatch is main session only.
+Class-2 platforms (Codex, Copilot, Gemini, Qoder) use the same dispatch prefix because they may not receive selected-task context through a pre-tool hook.
+Dispatch prompt starts with `Selected task: <task path from task.py selected>`. Read context: jsonl entries -> `prd.md` -> `design.md if present` -> `implement.md if present`.
+[/workflow-state:in_progress]
+
+<!-- Per-turn breadcrumb: shown while status='in_progress' when
+     codex.dispatch_mode=inline. Codex-only opt-in alternate to
+     [workflow-state:in_progress]. The main session edits code directly
+     instead of dispatching sub-agents. -->
+
+[workflow-state:in_progress-inline]
+Flow: `trellis-before-dev` -> edit -> `trellis-check` -> validation/evidence in `verify.md` -> learning decision -> commit (Phase 3.4) -> `task.py archive <task> --check` -> `/trellis:finish-work`.
+Execution boundary: edit only inside approved `prd.md`, `design.md`, `implement.md`, and Development Strategy Contract; stop and Return-to-Planning for scope, contract, gate, capability/runtime, Parent `contract_epoch`, Child boundary, selected-task fit, or non-implementation reviewer-gate changes.
+Do not dispatch implement/check sub-agents in inline mode.
+Read context: `prd.md` -> `design.md if present` -> `implement.md if present`, plus relevant spec/research loaded by skills.
+[/workflow-state:in_progress-inline]
+
+### Phase 3: Finish
+- 3.1 Quality verification `[required · repeatable]`
+- 3.2 Debug retrospective `[on demand]`
+- 3.3 Learning decision `[required · once]`
+- 3.4 Commit changes `[required · once]`
+- 3.5 Wrap-up reminder
+
+<!-- Per-turn breadcrumb: shown while status='completed'.
+     Currently DEAD in normal flow: cmd_archive writes status='completed' in
+     the same call that moves the task dir to archive/, so the selected-task
+     resolver loses the pointer and the hook never fires on archived tasks.
+     Block preserved for a future status-transition redesign (e.g. an
+     explicit in_progress→completed command). Edit through the same spec
+     channel as the live blocks. -->
+
+[workflow-state:completed]
+Code committed. Run `/trellis:finish-work`; if dirty, return to Phase 3.4 first.
+[/workflow-state:completed]
+
+### Rules
+
+1. Identify which Phase you're in, then continue from the next step there
+2. Run steps in order inside each Phase; `[required]` steps can't be skipped
+3. Phases can roll back (e.g., Execute reveals a prd defect → return to Plan to fix, then re-enter Execute)
+4. Steps tagged `[once]` are skipped if the output already exists; don't re-run
+5. Artifact presence informs the next step; missing `design.md` / `implement.md` is valid for lightweight tasks and incomplete planning for complex tasks.
+6. Return-to-Planning triggers must refresh affected artifacts, gates, fingerprints, and explicit execution approval before Execution resumes.
+
+### Active Task Routing
+
+When a user request matches one of these intents inside a selected task, route first, then load the detailed phase step if needed.
+
+[Claude Code, Cursor, OpenCode, codex-sub-agent, Kiro, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+- Planning or unclear requirements -> `trellis-brainstorm`.
+- `in_progress` implementation/check -> dispatch `trellis-implement` / `trellis-check`.
+- Repeated debugging -> `trellis-break-loop`; spec updates -> `trellis-update-spec`.
+
+[/Claude Code, Cursor, OpenCode, codex-sub-agent, Kiro, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+[codex-inline, Kilo, Antigravity, Windsurf]
+
+- Planning or unclear requirements -> `trellis-brainstorm`.
+- Before editing -> `trellis-before-dev`; after editing -> `trellis-check`.
+- Repeated debugging -> `trellis-break-loop`; spec updates -> `trellis-update-spec`.
+
+[/codex-inline, Kilo, Antigravity, Windsurf]
+
+### Guardrails
+
+- Task creation approval is not implementation approval; implementation waits for passing `task.py start-execution <task> --check`, explicit user execution approval, and `task.py start-execution <task> --approved`.
+- Planning stops at `task.py start-execution <task> --check` plus an explicit execution-approval request. Planning may not perform implementation edits, mutate execution status, start child execution, integrate children, archive, or claim completion.
+- Ordinary conversational confirmations such as "confirm", "agree", "ok", or "start" are not execution authorization unless they answer the explicit execution-approval prompt after a passing `--check`.
+- PRD-only is valid for lightweight tasks; complex tasks need `design.md` + `implement.md`.
+- Planning must be persisted to task artifacts; checks must run before reporting completion.
+- `verify.md` is the human-readable evidence center. `task.json.quality_gate_results` stores compact machine-checkable gate summaries and references only.
+
+### Return-to-Planning
+
+Return to Planning when continued execution would change the approved contract: PRD scope or acceptance criteria, design boundary, dependency, rollback, or validation strategy, Development Strategy Contract, `quality_gates`, `verification_profile`, capability/runtime assumptions, Parent `contract_epoch`, Child boundary, selected-task fit, or a non-implementation reviewer-gate root cause.
+
+When returning to Planning, update the affected planning artifacts, refresh gate records or fingerprints that depend on them, rerun `task.py start-execution <task> --check`, report the new task and contract/fingerprint context, and ask for explicit execution approval again before `--approved`.
+
+Implementation defects inside the already approved contract route back to Execution, not Planning. Validation environment blockers stay in Verification / Review until the environment or evidence path is resolved. Repeated same-issue review/gate loops require escalation to the user instead of silent retries.
+
+### Loading Step Detail
+
+At each step, run this to fetch detailed guidance:
+
+```bash
+python3 ./.trellis/scripts/get_context.py --mode phase --step <step>
+# e.g. python3 ./.trellis/scripts/get_context.py --mode phase --step 1.1
+```
+
+---
+
+## Phase 1: Plan
+
+Goal: classify the request, get task-creation consent when a task is needed, and produce the planning artifacts required before implementation.
+
+#### 1.0 Create task `[required · once]`
+
+Create the task directory only after task-creation consent. The command sets status to `planning`, writes `task.json`, and creates a default `prd.md`. It does not select the task, start execution, or approve execution:
+
+```bash
+python3 ./.trellis/scripts/task.py create "<task title>" --slug <name>
+```
+
+`--slug` is the human-readable name only. Do **not** include the `MM-DD-` date prefix; `task.py create` adds that prefix automatically.
+
+For task trees, create the parent task first and then create each child with `--parent <parent-dir>`. Do not start the parent just because children exist; start the child that owns the next independently verifiable deliverable.
+
+After creation, select the task only when the user has chosen it for this live session:
+
+```bash
+python3 ./.trellis/scripts/task.py select <task-dir>
+```
+
+After selection, the per-turn breadcrumb switches to `[workflow-state:planning]`, telling the AI to stay in planning.
+
+Run only `create` and, when appropriate, `select` here. Do not run `start-execution --approved` until step 1.4 passes its non-mutating `--check` and the user gives explicit execution approval.
+
+Skip when the user has already explicitly selected an appropriate task with `python3 ./.trellis/scripts/task.py select <task>`.
+
+#### 1.1 Requirement exploration `[required · repeatable]`
+
+Load the `trellis-brainstorm` skill and explore requirements interactively with the user per the skill's guidance.
+
+The brainstorm skill will guide you to:
+- Ask one question at a time
+- Prefer researching over asking the user
+- Prefer offering options over open-ended questions
+- Update `prd.md` immediately after each user answer
+- Split large scopes into a parent task plus child tasks when the deliverables can be verified independently
+- Keep `prd.md` focused on requirements and acceptance criteria
+- For complex tasks, produce `design.md` and `implement.md` before implementation starts
+
+When considering a parent/child split:
+- Use a parent task when one request contains several independently verifiable deliverables.
+- Parent tasks own source requirements, child-task mapping, cross-child acceptance criteria, and final integration review.
+- Child tasks own actual deliverables that can be planned, implemented, checked, and archived independently.
+- Parent/child structure is not a dependency system. If child B depends on child A, write that ordering in child B's `prd.md` / `implement.md`.
+- Start the child task that owns the next deliverable. Do not start the parent unless the parent itself has direct implementation work.
+
+Return to this step whenever requirements change and revise the relevant artifact.
+
+#### 1.2 Research `[optional · repeatable]`
+
+Research can happen at any time during requirement exploration. It isn't limited to local code — you can use any available tool (MCP servers, skills, web search, etc.) to look up external information, including third-party library docs, industry practices, API references, etc.
+
+[Claude Code, Cursor, OpenCode, codex-sub-agent, Kiro, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+Spawn the research sub-agent:
+
+- **Agent type**: `trellis-research`
+- **Task description**: Research <specific question>
+- **Key requirement**: Research output MUST be persisted to `{TASK_DIR}/research/`
+
+[/Claude Code, Cursor, OpenCode, codex-sub-agent, Kiro, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+[codex-inline, Kilo, Antigravity, Windsurf]
+
+Do the research in the main session directly and write findings into `{TASK_DIR}/research/`. (For `codex-inline` this avoids the `fork_turns="none"` isolation that prevents `trellis-research` sub-agents from resolving the selected task path.)
+
+[/codex-inline, Kilo, Antigravity, Windsurf]
+
+**Retrieval during research**:
+- Use `python3 ./.trellis/scripts/search_artifacts.py --query "<topic>" --json` to find durable Trellis specs, prior tasks, research, verification notes, and workspace journals before re-discovering framework context.
+- Use `codebase-retrieval` evidence levels for source-code questions: adapter output is candidate evidence until current source, Git, or validation confirms it.
+- Persist useful exploratory retrieval chains, adapter availability, and competing hypotheses under `{TASK_DIR}/research/`.
+
+**Research artifact conventions**:
+- One file per research topic (e.g. `research/auth-library-comparison.md`)
+- Record third-party library usage examples, API references, version constraints in files
+- Note relevant spec file paths you discovered for later reference
+- Optional reusable-research frontmatter can make findings easier to rediscover:
+  ```markdown
+  ---
+  doc_type: research
+  status: active
+  confidence: medium
+  scope: authentication
+  related_files:
+    - src/auth/login.ts
+  ---
+  ```
+- Recommended reusable-research sections: Quick Answer, Key Evidence, Details, Risks / Open Questions, Next Steps
+- Evidence claims should include file paths, commands, URLs, or validation output when available
+
+Brainstorm and research can interleave freely — pause to research a technical question, then return to talk with the user.
+
+**Key principle**: Research output must be written to files, not left only in the chat. Conversations get compacted; files don't.
+
+#### 1.3 Configure context `[required · once]`
+
+[Claude Code, Cursor, OpenCode, codex-sub-agent, Kiro, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+Curate `implement.jsonl` and `check.jsonl` so the Phase 2 sub-agents get the right spec/research context. These files were seeded on `task create` with a single self-describing `_example` line; your job here is to fill in real entries.
+
+**Location**: `{TASK_DIR}/implement.jsonl` and `{TASK_DIR}/check.jsonl` (already exist).
+
+**Format**: one JSON object per line — `{"file": "<path>", "reason": "<why>"}`. Paths are repo-root relative.
+
+**What to put in**:
+- **Spec files** — `.trellis/spec/<package>/<layer>/index.md` and any specific guideline files (`error-handling.md`, `conventions.md`, etc.) relevant to this task
+- **Research files** — `{TASK_DIR}/research/*.md` that the sub-agent will need to consult
+
+**What NOT to put in**:
+- Code files (`src/**`, `packages/**/*.ts`, etc.) — those are read by the sub-agent during implementation, not pre-registered here
+- Files you're about to modify — same reason
+
+**Split between the two files**:
+- `implement.jsonl` → specs + research the implement sub-agent needs to write code correctly
+- `check.jsonl` → specs for the check sub-agent (quality guidelines, check conventions, same research if needed)
+
+These manifests do not replace `implement.md`. `implement.md` is the human-readable execution plan for a complex task; jsonl files only list context files to inject or load.
+
+**How to discover relevant specs**:
+
+```bash
+python3 ./.trellis/scripts/get_context.py --mode packages
+```
+
+Lists every package + its spec layers with paths. Pick the entries that match this task's domain.
+
+Use artifact search when prior task/research evidence is likely relevant:
+
+```bash
+python3 ./.trellis/scripts/search_artifacts.py --query "<topic>" --json
+```
+
+Add any reusable `{TASK_DIR}/research/*.md` files you discovered when sub-agents need them.
+
+**How to append entries**:
+
+Either edit the jsonl file directly in your editor, or use:
+
+```bash
+python3 ./.trellis/scripts/task.py add-context "$TASK_DIR" implement "<path>" "<reason>"
+python3 ./.trellis/scripts/task.py add-context "$TASK_DIR" check "<path>" "<reason>"
+```
+
+Delete the seed `_example` line once real entries exist (optional — it's skipped automatically by consumers).
+
+Skip when: `implement.jsonl` and `check.jsonl` have agent-curated entries (the seed row alone doesn't count).
+
+[/Claude Code, Cursor, OpenCode, codex-sub-agent, Kiro, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+[codex-inline, Kilo, Antigravity, Windsurf]
+
+Skip this step. Context is loaded directly by the `trellis-before-dev` skill in Phase 2.
+
+[/codex-inline, Kilo, Antigravity, Windsurf]
+
+#### 1.4 Execution gate `[required · once]`
+
+After artifact review, run the non-mutating execution preflight:
+
+```bash
+python3 ./.trellis/scripts/task.py start-execution <task-dir> --check
+```
+
+For lightweight tasks, `prd.md` can be enough. For complex tasks, `prd.md`, `design.md`, and `implement.md` must exist and be reviewed before execution approval. On sub-agent-capable platforms, curate jsonl manifests when extra spec or research context is needed; seed-only manifests are tolerated by consumers.
+
+If `--check` passes, report that artifact gates are ready and include the task name/path plus current contract/fingerprint context. Then ask the user for explicit execution approval and state that approval permits `task.py start-execution <task-dir> --approved`. Ordinary agreement such as "confirm", "agree", "ok", or "start" before this preflight is not execution approval.
+
+Only after the user approves execution in that context, run:
+
+```bash
+python3 ./.trellis/scripts/task.py start-execution <task-dir> --approved
+```
+
+After this command succeeds, the breadcrumb switches to `[workflow-state:in_progress]`, and the rest of Phase 2 / 3 follows.
+
+#### 1.5 Completion criteria
+
+| Condition | Required |
+|------|:---:|
+| `prd.md` exists | ✅ |
+| `task.py start-execution <task> --check` passes | ✅ |
+| User explicitly approves execution after the passing preflight is reported | ✅ |
+| `task.py start-execution <task> --approved` has been run (status = in_progress) | ✅ |
+| `research/` has artifacts (complex tasks) | recommended |
+| `design.md` exists (complex tasks) | ✅ |
+| `implement.md` exists (complex tasks) | ✅ |
+
+[Claude Code, Cursor, OpenCode, codex-sub-agent, Kiro, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+| `implement.jsonl` / `check.jsonl` curated when extra spec or research context is needed | recommended |
+
+[/Claude Code, Cursor, OpenCode, codex-sub-agent, Kiro, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+---
+
+## Phase 2: Execute
+
+Goal: turn reviewed planning artifacts into code that passes quality checks.
+
+Execution is bounded by the approved `prd.md`, `design.md`, `implement.md`, and Development Strategy Contract. Follow the contract's `execution_mode`, `verification_profile`, and `quality_gates`; do not global-reclassify the request, auto-switch selected tasks, auto-create new task scope, or edit planning artifacts to change scope, design, or contract while pretending execution is still approved.
+
+Stop Execution and Return-to-Planning for PRD scope/acceptance changes, design boundary/dependency/rollback/validation changes, Development Strategy Contract changes, quality gate changes, capability/runtime assumption changes, Parent `contract_epoch` changes, Child boundary changes, selected-task scope mismatch, or non-implementation reviewer-gate failures. Refresh gates/fingerprints and get explicit approval before continuing.
+
+Implementation defects inside the approved contract remain Execution work: fix them in Phase 2, update `verify.md` evidence, and re-run validation without changing the approved contract.
+
+#### 2.1 Implement `[required · repeatable]`
+
+Use retrieval layers before and during implementation when context is incomplete:
+- `python3 ./.trellis/scripts/search_artifacts.py --query "<topic>" --json` for durable Trellis specs, prior task artifacts, research, verification notes, and journals.
+- `codebase-retrieval` evidence levels for source claims: candidate -> corroborated candidate -> verified claim; unresolved or unavailable adapters must be reported instead of treated as proof.
+- Record exploratory chains in `{TASK_DIR}/research/` and final source/Git/test proof in `verify.md`.
+
+[Claude Code, Cursor, OpenCode, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+Spawn the implement sub-agent:
+
+- **Agent type**: `trellis-implement`
+- **Task description**: Implement the reviewed task artifacts, consulting materials under `{TASK_DIR}/research/`; finish by running project lint and type-check
+- **Dispatch prompt guard**: Tell the spawned agent it is already the `trellis-implement` sub-agent and must implement directly, not spawn another `trellis-implement` / `trellis-check`.
+
+The platform hook/plugin auto-handles:
+- Reads `implement.jsonl` and injects referenced spec/research files into the agent prompt
+- Injects `prd.md`, `design.md` if present, and `implement.md` if present
+
+[/Claude Code, Cursor, OpenCode, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+[codex-sub-agent]
+
+Spawn the implement sub-agent:
+
+- **Agent type**: `trellis-implement`
+- **Task description**: Implement the reviewed task artifacts, consulting materials under `{TASK_DIR}/research/`; finish by running project lint and type-check
+- **Dispatch prompt guard**: The prompt MUST start with `Selected task: <task path>`, then explicitly say the spawned agent is already `trellis-implement` and must implement directly without spawning another `trellis-implement` / `trellis-check`.
+
+The Codex sub-agent definition auto-handles the context load requirement:
+- Resolves the selected task with `task.py selected --source`, then reads `prd.md`, `design.md` if present, and `implement.md` if present
+- Reads `implement.jsonl` and requires the agent to load each referenced spec/research file before coding
+
+[/codex-sub-agent]
+
+[Kiro]
+
+Spawn the implement sub-agent:
+
+- **Agent type**: `trellis-implement`
+- **Task description**: Implement the reviewed task artifacts, consulting materials under `{TASK_DIR}/research/`; finish by running project lint and type-check
+- **Dispatch prompt guard**: Tell the spawned agent it is already the `trellis-implement` sub-agent and must implement directly, not spawn another `trellis-implement` / `trellis-check`.
+
+The platform prelude auto-handles the context load requirement:
+- Reads `implement.jsonl` and injects referenced spec/research files into the agent prompt
+- Injects `prd.md`, `design.md` if present, and `implement.md` if present
+
+[/Kiro]
+
+[codex-inline, Kilo, Antigravity, Windsurf]
+
+1. Load the `trellis-before-dev` skill to read project guidelines
+2. Read `{TASK_DIR}/prd.md`, then `design.md` if present, then `implement.md` if present
+3. Consult materials under `{TASK_DIR}/research/`
+4. Implement the code per reviewed artifacts
+5. Run project lint and type-check
+
+[/codex-inline, Kilo, Antigravity, Windsurf]
+
+#### 2.2 Quality check `[required · repeatable]`
+
+[Claude Code, Cursor, OpenCode, codex-sub-agent, Kiro, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+Spawn the check sub-agent:
+
+- **Agent type**: `trellis-check`
+- **Task description**: Review all code changes against specs and task artifacts; fix implementation defects only inside the approved contract; route contract-changing findings instead of silently expanding scope; ensure lint and type-check pass
+- **Dispatch prompt guard**: Tell the spawned agent it is already the `trellis-check` sub-agent and must review/fix only inside the approved contract, not spawn another `trellis-check` / `trellis-implement`.
+
+The check agent's job:
+- Review code changes against specs
+- Review code changes against `prd.md`, `design.md` if present, and `implement.md` if present
+- Fix implementation defects only when they stay inside the approved contract
+- Route requirement, design, contract, scope, gate, capability, runtime, Parent `contract_epoch`, or Child boundary defects to Return-to-Planning
+- Run lint and typecheck to verify
+
+[/Claude Code, Cursor, OpenCode, codex-sub-agent, Kiro, Gemini, Qoder, CodeBuddy, Copilot, Droid, Pi]
+
+[codex-inline, Kilo, Antigravity, Windsurf]
+
+Load the `trellis-check` skill and verify the code per its guidance:
+- Spec compliance
+- lint / type-check / tests
+- Cross-layer consistency (when changes span layers)
+
+If implementation defects are found inside the approved contract -> fix in Execution -> re-check. If findings change requirements, design, contract, scope, gates, or capability/runtime assumptions -> Return-to-Planning.
+
+[/codex-inline, Kilo, Antigravity, Windsurf]
+
+#### 2.3 Rollback `[on demand]`
+
+- `check` reveals a PRD, design, contract, scope, gate, capability/runtime, Parent `contract_epoch`, or Child boundary defect → Return-to-Planning, refresh gates/fingerprints, and get explicit approval again
+- Implementation went wrong → revert code, redo 2.1
+- Need more research → research (same as Phase 1.2), write findings into `research/`
+
+---
+
+## Phase 3: Finish
+
+Goal: ensure code quality, capture lessons, record the work.
+
+#### Evidence scoring (retrieval-pack, explicit)
+
+For **complex** tasks, optionally before 3.1 when you have collected retrieval evidence JSON (e.g. after Phase 1.2 smart-search manifests in `{TASK}/research/smart-search/`):
+
+`python3 ./.trellis/scripts/get_context.py --mode retrieval-pack --json --input <path-to-evidence.json>`
+
+Use `contextPack.selected` / `scoredEvidence` to order citations in `verify.md`. Do **not** expect scoring in default `get_context --json`.
+
+#### 3.1 Quality verification `[required · repeatable]`
+
+Verification / Review is evidence and judgment, not a hidden implementation loop. Load the `trellis-check` skill or agent and do a final review:
+- Spec compliance
+- lint / type-check / tests
+- Cross-layer consistency (when changes span layers)
+- Retrieval evidence: final claims must cite current source, Git, or validation proof; unresolved adapter or artifact-search gaps belong in `verify.md`
+- Reviewer gate evidence via `task.py record-gate <task> <transition> <gate> <PASS|FAIL|SKIPPED>` when a configured gate applies
+- Human-readable validation, review, and acceptance evidence in `verify.md`
+
+Do not silently implement fixes, expand scope, or edit planning artifacts during Verification / Review. Review-found implementation defects inside the approved contract route back to Execution. Requirement, design, contract, scope, gate-configuration, capability/runtime, Parent `contract_epoch`, or Child boundary defects route back to Planning. Validation environment blockers stay in Verification / Review with explicit blocker evidence.
+
+`verify.md` is the evidence center for humans. `task.json.quality_gate_results` should contain only compact machine-checkable results, fingerprints, timestamps, reviewer ids, evidence references, and root-cause metadata.
+
+#### 3.2 Debug retrospective `[on demand]`
+
+If this task involved repeated debugging (the same issue was fixed multiple times), load the `trellis-break-loop` skill to:
+- Classify the root cause
+- Explain why earlier fixes failed
+- Propose prevention
+
+The goal is to capture debugging lessons so the same class of issue doesn't recur.
+
+#### 3.3 Learning decision `[required · once]`
+
+Review whether this task produced durable learning worth recording:
+- repeated failure loops or debugging lessons;
+- requirement drift, architecture decisions, reusable conventions, or toolchain pitfalls;
+- new project-local patterns that should affect future work.
+
+If durable learning exists, load `trellis-update-spec` and update `.trellis/spec/` or write a focused `retrospective.md`, then link that evidence from `verify.md`.
+
+If no durable learning exists, write an explicit `No durable learning` decision in `verify.md`. Do not run a spec update only to satisfy ceremony.
+
+Before archive, `verify.md` must contain validation evidence, final acceptance evidence, gate/review references when applicable, and the durable-learning decision. Parent tasks with children must also include final integration evidence.
+
+#### 3.4 Commit changes `[required · once]`
+
+The AI drives a batched commit of this task's code changes so `/finish-work` can run cleanly afterwards. Goal: produce work commits FIRST, then bookkeeping (archive + journal) commits land after — never interleaved.
+
+**Step-by-step**:
+
+1. **Inspect dirty state**:
+   ```bash
+   git status --porcelain
+   ```
+   Snapshot every dirty path. If the working tree is clean, skip to 3.5.
+
+2. **Learn commit style** from recent history (so drafted messages blend in):
+   ```bash
+   git log --oneline -5
+   ```
+   Note the prefix convention (`feat:` / `fix:` / `chore:` / `docs:` ...), language (中文/English), and length style.
+
+3. **Classify dirty files into two groups**:
+   - **AI-edited this session** — files you wrote/edited via Edit/Write/Bash tool calls in this session. You know what changed and why.
+   - **Unrecognized** — dirty files you did NOT touch this session (could be the user's manual edits, leftover WIP from a previous session, or unrelated work). Do NOT silently include these.
+
+4. **Draft a commit plan**. Group AI-edited files into logical commits (1 commit per coherent change unit, not 1 commit per file). Each entry: `<commit message>` + file list. List unrecognized files separately at the bottom.
+
+5. **Present the plan once, ask for one-shot confirmation**. Format:
+   ```
+   Proposed commits (in order):
+     1. <message>
+        - <file>
+        - <file>
+     2. <message>
+        - <file>
+
+   Unrecognized dirty files (NOT in any commit — confirm include/exclude):
+     - <file>
+     - <file>
+
+   Reply 'ok' / '行' to execute. Reply with edits, or '我自己来' / 'manual' to abort.
+   ```
+
+6. **On confirmation**: run `git add <files>` + `git commit -m "<msg>"` for each batch in order. Do not amend. Do not push.
+
+7. **On rejection** (user replies "不行" / "我自己来" / "manual" / any pushback on the plan): stop. Do not attempt a second plan. The user will commit by hand; you skip ahead to 3.5 once they confirm.
+
+**Rules**:
+- No `git commit --amend` anywhere — three-stage three-commit flow (work commits → archive commit → journal commit).
+- Never push to remote in this step.
+- If the user wants different message wording but accepts the file grouping, edit the message and re-confirm once — but if they reject the grouping, exit to manual mode.
+- The batched plan is one prompt; do not prompt per commit.
+
+#### 3.5 Archive / Learning
+
+Run `task.py archive <task> --check` before the real archive. The check must pass without moving files, changing status, clearing selected-task state, staging or committing, or running hooks.
+
+Real completion is only `task.py archive <task>`. It writes `status=completed`, moves the task to `archive/<YYYY-MM>/`, clears selected-task pointers for that task, and runs archive hooks. Passing validation, user acceptance, reviewer gates, or writing `verify.md` alone does not complete the task.
+
+Archive readiness by mode:
+
+| Mode | Archive readiness |
+|---|---|
+| No Task | No archive; upgrade to a durable task mode before archive is possible. |
+| Micro-Grill | No archive unless upgraded into Lite, Full, or Parent/Child. |
+| Lite | `verify.md` has validation evidence, final acceptance evidence, and durable-learning decision evidence. |
+| Full | Lite evidence plus required completion gates, fresh fingerprints, and no unresolved required `FAIL` gates. |
+| Child | Lite/Full evidence plus Parent task-map marks the Child `integrated` or `cancelled`; integrated Children include `handoff.md`. |
+| Parent | Full evidence plus every Child is `integrated` or `cancelled`, with final integration evidence. |
+
+Archive / Learning is terminal. After archive, do not silently mutate archived task artifacts; follow-up work requires a new task unless the user explicitly approves an archive amendment.
+
+If archive is not being run in this session, report the passing or failing archive check and the remaining evidence gaps.
+
+---
+
+## Release readiness and release execution
+
+Package and CLI releases use **two tasks** (or two explicit phases with separate evidence). Readiness is non-mutating; execution requires explicit user approval before any remote mutation.
+
+### When to split
+
+| Task type | Purpose | Remote mutations |
+| --- | --- | --- |
+| **Release readiness** | Version recommendation, changelog, manifest notes, build/test/typecheck, pack dry-run, prepublish blockers | **None** — no publish, tag, push, or version bump |
+| **Release execution** | Preflight immediately before publish; run approved `release.js` / `npm publish` / tag / push; post-publish smoke | **Only after explicit user approval** in the execution task |
+
+Templates: `.trellis/tasks/templates/release-readiness/` and `.trellis/tasks/templates/release-execution/`. Copy `prd.md`, `design.md`, and `implement.md` into a new task directory; use `handoff-template.md` as the outline for `handoff.md`.
+
+### Release status vocabulary (handoff / verify)
+
+These labels describe **release posture** in `handoff.md` and `verify.md`. They are not `task.json.status` values (contract epoch 1).
+
+| Label | Typical task |
+| --- | --- |
+| **ready to publish** | Readiness — evidence complete; operator may proceed after approval |
+| **not published** | Readiness or aborted execution — no remote release in this task |
+| **published** | Execution — registry and/or git remote actions completed |
+| **blocked** | Required gate failed; do not publish until fixed or waived |
+| **waived** | User explicitly accepted a known blocker (name approver + rationale) |
+| **deferred** | Publish postponed by user (e.g. wait for dogfood / another child) |
+
+Readiness `handoff.md` must include **Ready to publish**, **Not published**, and **Blockers**. Execution `handoff.md` must include **Published** (or **Not published** if aborted) and **Publish approval evidence** in `verify.md`.
+
+### Standard evidence (readiness `verify.md`)
+
+- Version recommendation (current → proposed; channel; npm dist-tag expectation)
+- Changelog / release notes (draft)
+- Migration manifest (id or none)
+- Build evidence, test evidence, pack dry-run evidence
+- Prepublish blockers with disposition: `blocking` / `fixed` / `waived` / `deferred`
+
+Dry-run examples for `@blxzer/cursor-trellis` (from `packages/cli`): `check-manifest-continuity.js`, `release-preflight.js check-versions`, `publish-plan`, `verify-packed-cli`, `npm pack --dry-run`. Do not run `release.js`, `npm publish`, or `git push` during readiness.
+
+### Execution approval gate
+
+Before `release.js`, `npm publish`, `git tag`, `git push`, or GitHub release creation, stop and obtain **explicit user approval**. Record in execution `verify.md` under **Publish approval evidence** (approved version, dist-tag, and which remote actions are allowed). Passing preflight is not approval.
+
+See also `.trellis/spec/Trellis/cli/release-and-vendor.md`.
+
+---
+
+## Customizing Trellis (for forks)
+
+This section is for developers who want to modify the Trellis workflow itself. All customization is done by editing this file; the scripts are parsers only.
+
+### Changing what a step means
+
+Edit the corresponding step's walkthrough body in the Phase 1 / 2 / 3 sections above. Critical invariants:
+- No selected task must use framework/dashboard routing, then triage and ask for task-creation consent before creating a Trellis task.
+- Planning must distinguish lightweight PRD-only tasks from complex tasks that require `prd.md`, `design.md`, and `implement.md` before start.
+- Every required execution path must keep the Phase 3.4 commit reminder reachable before `/trellis:finish-work`.
+
+All tag blocks live in the `## Phase Index` section above, immediately after each phase summary:
+
+| Scope | Corresponding tag |
+|---|---|
+| No selected task (before Phase 1) | `[workflow-state:no_task]` (after the Phase Index ASCII art) |
+| All of Phase 1 (task created → ready for implementation) | `[workflow-state:planning]` (after Phase 1 summary) |
+| Codex inline Phase 1 | `[workflow-state:planning-inline]` |
+| Phase 2 + Phase 3.1–3.4 (implementation + check + wrap-up) | `[workflow-state:in_progress]` (after Phase 2 summary) |
+| Codex inline Phase 2 + Phase 3.1–3.4 | `[workflow-state:in_progress-inline]` |
+| After Phase 3.5 (archived) | `[workflow-state:completed]` (after Phase 3 summary; **currently DEAD**) |
+
+### Changing the per-turn prompt text
+
+Directly edit the body of the corresponding `[workflow-state:STATUS]` block. After editing, run `trellis update` (if you're a template maintainer) or restart your AI session (if you're customizing your own project) — no script changes required.
+
+### Adding a custom status
+
+Add a new block:
+
+```
+[workflow-state:my-status]
+your per-turn prompt text
+[/workflow-state:my-status]
+```
+
+Constraints:
+- STATUS charset: `[A-Za-z0-9_-]+` (underscores and hyphens allowed, e.g. `in-review`, `blocked-by-team`)
+- A lifecycle hook must write `task.json.status` to your custom value, otherwise the tag is never read
+- Lifecycle hooks live in `task.json.hooks.after_*` and bind to one of `after_create / after_start / after_finish / after_archive`
+
+### Adding a lifecycle hook
+
+Add a `hooks` field to your `task.json`:
+
+```json
+{
+  "hooks": {
+    "after_archive": [
+      "your-script-or-command-here"
+    ]
+  }
+}
+```
+
+Supported events: `after_create / after_start / after_archive`. Use `after_archive` for "task is done" notifications. Historical `after_finish` hooks are not part of the selected-task workflow because `task.py finish` no longer exists.
+
+### Full contract
+
+For the workflow state machine's runtime contract, the locations of all status writers, pseudo-statuses (`no_task` / `stale_<source_type>`), the hook reachability matrix, and other deep details, see:
+
+- `.trellis/spec/cli/backend/workflow-state-contract.md` — runtime contract + writer table + test invariants
+- `.trellis/scripts/inject-workflow-state.py` — actual parser (reads workflow.md only, no embedded text)