wtf-p 0.3.0 → 0.5.0

package/core/write-the-f-paper/references/orchestrator-pattern.md

@@ -0,0 +1,167 @@
+# Thin Orchestrator Pattern
+
+All refactored WTF-P commands follow this pattern. Orchestrators are lightweight routers — they validate, load context, resolve models, spawn agents, and handle returns. Heavy work lives in agents.
+
+## Why Thin Orchestrators
+
+- **Fresh context**: Agents start with clean context windows, no orchestrator overhead
+- **Peak quality**: Full context budget goes to the actual task, not workflow machinery
+- **Independent verification**: Quality agents (plan-checker, argument-verifier) validate without seeing the generation process
+- **Composability**: Same agent can serve multiple orchestrators
+
+## Command Frontmatter
+
+```yaml
+---
+name: wtfp:[command-name]
+description: [Brief description]
+argument-hint: "[what user provides]"
+allowed-tools:
+  - Read
+  - Bash
+  - Write
+  - Glob
+  - Grep
+  - Task
+  - AskUserQuestion
+---
+```
+
+## execution_context
+
+References to workflow/reference/template files the orchestrator needs:
+
+```markdown
+<execution_context>
+@~/.claude/write-the-f-paper/workflows/[workflow].md
+@~/.claude/write-the-f-paper/references/[reference].md
+@~/.claude/write-the-f-paper/templates/[template].md
+</execution_context>
+```
+
+## Objective Section
+
+```markdown
+<objective>
+[Clear statement of what this command does]
+
+**Orchestrator role:** Validate → Load context → Resolve model profile → Spawn agent → Handle returns → Present results
+
+**Why subagents:** [Rationale — typically context preservation and independent quality checks]
+</objective>
+```
+
+## Process Structure
+
+Every orchestrator follows these phases:
+
+### 1. Validate Environment and Resolve Model Profile
+
+```bash
+ls .planning/ 2>/dev/null
+MODEL_PROFILE=$(cat .planning/config.json 2>/dev/null | \
+  grep -o '"model_profile"[[:space:]]*:[[:space:]]*"[^"]*"' | \
+  grep -o '"[^"]*"$' | tr -d '"' || echo "balanced")
+```
+
+Include per-command model lookup table (see agent-model-matrix.md).
+
+### 2. Load Context Files
+
+```bash
+STATE_CONTENT=$(cat .planning/STATE.md)
+PROJECT_CONTENT=$(cat .planning/PROJECT.md)
+CONTEXT_CONTENT=$(cat "${SECTION_DIR}"/*-CONTEXT.md 2>/dev/null)
+```
+
+CRITICAL: Store CONTEXT_CONTENT now. Pass to ALL downstream agents.
+
+### 3. Parse Arguments / Validate Section
+
+Command-specific validation (section exists, plan exists, dependencies met).
+
+### 4. Spawn Primary Agent
+
+```
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+ WTF-P ► [ACTION] SECTION {X}
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+```
+
+Build prompt with inlined content, then spawn:
+
+```
+Task(
+  prompt="First, read ~/.claude/agents/wtfp/[agent].md for your role and instructions.\n\n" + filled_prompt,
+  subagent_type="general-purpose",
+  model="{resolved_model}",
+  description="[Action] Section {X}"
+)
+```
+
+The filled prompt includes:
+- `<[context_type]>` with STATE, PROJECT, ROADMAP, structure docs, prior summaries
+- `<user_decisions>` with CONTEXT_CONTENT from CONTEXT.md
+- `<output>` with target file paths
+
+### 5. Handle Agent Return
+
+Agents return structured headers that orchestrators parse:
+
+| Return Header | Meaning | Orchestrator Action |
+|--------------|---------|-------------------|
+| `## [ACTION] COMPLETE` | Success | Proceed to quality agent or next step |
+| `## CHECKPOINT REACHED` | Needs user input | Present to user, wait, resume |
+| `## [ACTION] BLOCKED` | Cannot proceed | Present blocker, ask user |
+
+### 6. Spawn Quality Agent (if applicable)
+
+Commands with quality loops spawn a second agent to verify:
+
+| Command | Quality Agent | Checks |
+|---------|--------------|--------|
+| plan-section | plan-checker | 7 dimensions: argument, citation, word budget, outline, CONTEXT.md fidelity, style, task completeness |
+| write-section | argument-verifier | Goal-backward: does output achieve section purpose? |
+
+### 7. Present Final Status
+
+```markdown
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+ WTF-P ► [STATUS] ✓
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+[Summary]
+
+───────────────────────────────────────────
+
+## ▶ Next Up
+
+**[Next action]** — [description]
+
+`/wtfp:[next-command] [args]`
+
+<sub>`/clear` first → fresh context window</sub>
+
+───────────────────────────────────────────
+```
+
+## Agent Prompt Construction
+
+Orchestrators build agent prompts by inlining content, not referencing files:
+
+```
+prompt = "First, read ~/.claude/agents/wtfp/{agent}.md for your role.\n\n"
+       + "<planning_context>\n" + STATE + PROJECT + ROADMAP + structure_docs + "\n</planning_context>\n"
+       + "<user_decisions>\n" + CONTEXT_CONTENT + "\n</user_decisions>\n"
+       + "<output>\n" + target_path + "\n</output>"
+```
+
+The agent file reference (`read ~/.claude/agents/...`) provides role instructions. The inlined content provides runtime context.
+
+## Anti-Patterns
+
+- **Fat orchestrator**: Orchestrator doing the actual work (writing, planning) instead of spawning agents
+- **Missing context injection**: Forgetting to pass CONTEXT.md to agents
+- **File references instead of inlining**: Agents can't read files the orchestrator could — inline the content
+- **Ignoring structured returns**: Not parsing the agent's return header for status routing
+- **Skipping quality loop**: Not spawning plan-checker or argument-verifier when config enables them

package/core/write-the-f-paper/references/planning-config.md

@@ -0,0 +1,135 @@
+# Planning Configuration Reference
+
+Complete documentation of every key in `.planning/config.json`. This is the v0.5.0 schema — the single source of truth for all orchestrator and agent configuration.
+
+Template location: `core/write-the-f-paper/templates/config.json`
+
+## Top-Level Settings
+
+| Key | Type | Default | Description | Read By |
+|-----|------|---------|-------------|---------|
+| `mode` | string | `"interactive"` | Execution mode. `"interactive"`: confirmations at gates. `"yolo"`: skip all non-safety gates. | All orchestrators |
+| `depth` | string | `"standard"` | Planning depth. `"quick"`: fewer plans, faster. `"standard"`: balanced. `"thorough"`: max detail. | section-planner, plan-checker |
+| `document_type` | string | `"paper"` | Document type. `"paper"`, `"thesis"`, `"proposal"`, `"report"`. Affects templates and defaults. | new-paper, plan-section |
+| `output_format` | string | `"markdown"` | Output file format. `"markdown"` or `"latex"`. Determines how content files are written. | section-writer, prose-polisher |
+| `model_profile` | string | `"balanced"` | Model selection profile. `"quality"`, `"balanced"`, `"budget"`. See agent-model-matrix.md. | All orchestrators (model resolution) |
+| `venue_template` | string\|null | `null` | Venue-specific template name. `null` for generic. When set, loads venue constraints (word limits, formatting). | plan-section, write-section |
+
+## Gates
+
+Controls which confirmation prompts appear. When `true`, orchestrator pauses for user confirmation. When `false`, orchestrator proceeds automatically. Mode `"yolo"` overrides all non-safety gates to `false`.
+
+| Key | Type | Default | Description | Read By |
+|-----|------|---------|-------------|---------|
+| `gates.confirm_outline` | bool | `true` | Confirm before finalizing outline structure. | create-outline, execute-outline |
+| `gates.confirm_plan` | bool | `true` | Confirm before finalizing section plan. | plan-section |
+| `gates.confirm_write` | bool | `true` | Confirm before executing writing plan. | write-section |
+| `gates.confirm_review` | bool | `true` | Confirm before applying review suggestions. | review-section |
+| `gates.execute_next_section` | bool | `true` | Confirm before auto-advancing to next section. | write-section (section chaining) |
+| `gates.issues_review` | bool | `true` | Confirm before presenting issues for review. | review-section, check-refs |
+| `gates.confirm_transition` | bool | `true` | Confirm before transitioning between writing phases. | write-section (phase transitions) |
+| `gates.confirm_submission` | bool | `true` | Confirm before marking paper as submission-ready. | submit-milestone |
+
+### When to Change
+
+- **Solo iterating rapidly**: Set `mode: "yolo"` to skip all gates.
+- **Specific gate annoying**: Set that one gate to `false`, keep others.
+- **Submission safety**: Never set `confirm_submission` to `false` — it is the final human check.
+
+## Writing
+
+Controls how Claude assists with writing. Preserved from v0.4.0.
+
+| Key | Type | Default | Description | Read By |
+|-----|------|---------|-------------|---------|
+| `writing.claude_mode` | string | `"co-author"` | Claude's writing role. `"co-author"`: Claude drafts prose. `"scaffold"`: Claude outlines, user writes. `"reviewer"`: User writes, Claude reviews. | section-writer, write-section |
+| `writing.citation_style` | string | `"apa"` | Citation format. `"apa"`, `"ieee"`, `"chicago"`, `"mla"`, `"vancouver"`. | citation-formatter, section-writer |
+| `writing.verify_on_complete` | bool | `true` | Run argument-verifier after each writing task completes. When `false`, skips verification. | write-section |
+
+## Workflow
+
+Controls which optional workflow steps run. Preserved from v0.4.0.
+
+| Key | Type | Default | Description | Read By |
+|-----|------|---------|-------------|---------|
+| `workflow.research` | bool | `true` | Run research-synthesizer before planning. When `false`, skip literature search. | plan-section |
+| `workflow.plan_check` | bool | `true` | Run plan-checker after plan creation. When `false`, skip quality validation. | plan-section |
+| `workflow.verifier` | bool | `true` | Run argument-verifier after writing. When `false`, skip goal-backward verification. | write-section |
+
+## Verification
+
+Controls which verification checks run during review. Preserved from v0.4.0.
+
+| Key | Type | Default | Description | Read By |
+|-----|------|---------|-------------|---------|
+| `verification.citation_check` | bool | `true` | Verify citations exist, are formatted correctly, and are cited in text. | check-refs, section-reviewer |
+| `verification.coherence_check` | bool | `true` | Check cross-section coherence, argument flow, terminology consistency. | section-reviewer, coherence-checker |
+| `verification.rubric_check` | bool | `true` | Evaluate against venue rubric (if venue_template set) or general quality rubric. | section-reviewer |
+
+## Planning
+
+Controls planning document behavior. New in v0.5.0.
+
+| Key | Type | Default | Description | Read By |
+|-----|------|---------|-------------|---------|
+| `planning.commit_docs` | bool | `true` | Include planning docs (SUMMARY.md, STATE.md) in git commits. When `false`, planning docs are gitignored. | All orchestrators (commit step) |
+| `planning.search_gitignored` | bool | `false` | Allow searching gitignored files for context. When `true`, agents can read files in .gitignore paths. | research-synthesizer, section-writer |
+
+## Parallelization
+
+Controls concurrent execution. Expanded in v0.5.0.
+
+| Key | Type | Default | Description | Read By |
+|-----|------|---------|-------------|---------|
+| `parallelization.enabled` | bool | `false` | Master switch for parallel execution. When `false`, all work is sequential. | All orchestrators |
+| `parallelization.plan_level` | bool | `true` | Allow parallel plan execution within a section (when `enabled` is `true`). | write-section |
+| `parallelization.task_level` | bool | `false` | Allow parallel task execution within a plan (when `enabled` is `true`). Experimental. | section-writer |
+| `parallelization.skip_checkpoints` | bool | `true` | Auto-approve checkpoints during parallel execution to avoid blocking. | section-writer |
+| `parallelization.max_concurrent_agents` | int | `3` | Maximum number of concurrent subagents. | All orchestrators (parallel mode) |
+| `parallelization.min_plans_for_parallel` | int | `2` | Minimum number of ready plans before parallel execution activates. | write-section |
+
+## Safety
+
+Controls safety confirmations that are never skipped by mode or gate overrides. Expanded in v0.5.0.
+
+| Key | Type | Default | Description | Read By |
+|-----|------|---------|-------------|---------|
+| `safety.always_confirm_destructive` | bool | `true` | Always confirm before destructive operations (deleting sections, resetting progress). Never skipped by `mode: "yolo"`. | All orchestrators |
+| `safety.always_confirm_external_services` | bool | `true` | Always confirm before calling external services (API calls, web searches). Never skipped by `mode: "yolo"`. | research-synthesizer, citation-retriever |
+| `safety.backup_before_major_edits` | bool | `true` | Create backup commit before major edits (full section rewrites, structural changes). | write-section, review-section |
+
+### Safety vs Gates
+
+Gates are convenience confirmations — skip them to go faster. Safety flags protect against data loss and unintended side effects. `mode: "yolo"` overrides gates but never overrides safety flags.
+
+## Git
+
+Controls git integration behavior. New in v0.5.0. See git-integration.md for full branching strategy documentation.
+
+| Key | Type | Default | Description | Read By |
+|-----|------|---------|-------------|---------|
+| `git.branching_strategy` | string | `"none"` | Branch strategy. `"none"`: current branch. `"section"`: branch per section. `"submission"`: branch per milestone. | All orchestrators (commit step) |
+| `git.section_branch_template` | string | `"wtfp/section-{section}-{slug}"` | Branch name template for `"section"` strategy. `{section}` = section number, `{slug}` = slugified section name. | plan-section, write-section |
+| `git.submission_branch_template` | string | `"wtfp/{milestone}-{slug}"` | Branch name template for `"submission"` strategy. `{milestone}` = milestone name, `{slug}` = slugified paper title. | submit-milestone |
+| `git.squash_on_merge` | bool | `false` | Squash commits when merging section/submission branches back. When `true`, per-task history is lost on merge. | write-section, submit-milestone |
+
+## Backward Compatibility
+
+The following keys are preserved unchanged from v0.4.0:
+
+- **`writing`** section: `claude_mode`, `citation_style`, `verify_on_complete` -- same keys, same defaults, same behavior.
+- **`verification`** section: `citation_check`, `coherence_check`, `rubric_check` -- same keys, same defaults, same behavior.
+- **`output_format`**: Same key, same default (`"markdown"`), same behavior.
+
+The following keys were renamed or replaced in v0.5.0:
+
+| v0.4.0 Key | v0.5.0 Key | Reason |
+|------------|------------|--------|
+| `gates.confirm_project` | Removed | Replaced by `gates.confirm_outline` |
+| `gates.confirm_sections` | Removed | Replaced by `gates.confirm_plan` |
+| `gates.confirm_roadmap` | Removed | No equivalent (roadmap is auto-generated) |
+| `gates.write_next_plan` | `gates.execute_next_section` | Renamed for clarity |
+| `parallelization.min_sections_for_parallel` | `parallelization.min_plans_for_parallel` | Renamed: plans not sections |
+| `safety.backup_before_major_edits` | Kept (+ `always_confirm_external_services` added) | Backward compat |
+
+Projects using v0.4.0 config files will work -- orchestrators default missing keys safely. The `new-paper` command generates the full v0.5.0 schema.

package/core/write-the-f-paper/references/ui-brand.md

@@ -0,0 +1,186 @@
+# WTF-P Visual Identity Reference
+
+Visual formatting conventions for all orchestrators and agents. Claude Code output is plain text — these symbols and patterns provide visual hierarchy without ANSI colors.
+
+## Stage Banners
+
+The `WTF-P ►` banner marks stage transitions in orchestrator output. Format:
+
+```
+WTF-P ► {STAGE_NAME} {optional_detail}
+```
+
+Rendered inside a box:
+
+```
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+ WTF-P ► {STAGE_NAME} {optional_detail}
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+```
+
+### Stage Names
+
+| Stage | Usage | Example |
+|-------|-------|---------|
+| INITIALIZING | Project setup, config load | `WTF-P ► INITIALIZING new-paper` |
+| PLANNING | Section plan creation | `WTF-P ► PLANNING SECTION 2` |
+| RESEARCHING | Literature search, gap analysis | `WTF-P ► RESEARCHING literature gaps` |
+| WRITING | Section draft execution | `WTF-P ► WRITING SECTION 1 PLAN 1` |
+| REVIEWING | Section review, quality checks | `WTF-P ► REVIEWING SECTION 3` |
+| POLISHING | Prose refinement, voice adjustment | `WTF-P ► POLISHING SECTION 1` |
+| VERIFYING | Argument verification, citation check | `WTF-P ► VERIFYING SECTION 2` |
+| CHECKING | Pre-flight checks, plan validation | `WTF-P ► CHECKING plan quality` |
+| COMPLETE | Stage finished successfully | `WTF-P ► COMPLETE ✓` |
+
+### Completion Banner
+
+```
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+ WTF-P ► COMPLETE ✓
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+[Summary of what was done]
+
+───────────────────────────────────────────
+
+## ▶ Next Up
+
+**[Next action]** — [description]
+
+`/wtfp:[next-command] [args]`
+
+<sub>`/clear` first → fresh context window</sub>
+
+───────────────────────────────────────────
+```
+
+## Checkpoint Boxes
+
+How checkpoints render to the user. Three types, same box format:
+
+```
+════════════════════════════════════════
+CHECKPOINT: {type}
+════════════════════════════════════════
+Task {X} of {Y}: {Name}
+
+{content — varies by checkpoint type}
+
+{resume signal}
+════════════════════════════════════════
+```
+
+### human-verify Box
+
+```
+════════════════════════════════════════
+CHECKPOINT: human-verify
+════════════════════════════════════════
+Task 3 of 5: Review introduction draft
+
+What was written:
+  Introduction opening paragraph (~200 words)
+
+How to verify:
+  1. Captures the problem accurately
+  2. Tone matches your voice
+  3. Hook is compelling
+
+Type "approved" or describe what to change.
+════════════════════════════════════════
+```
+
+### decision Box
+
+```
+════════════════════════════════════════
+CHECKPOINT: decision
+════════════════════════════════════════
+Task 2 of 4: Methodology framing
+
+Decision needed:
+  How to frame the methodology section
+
+Options:
+  [a] Emphasize novelty — highlights contribution, may invite scrutiny
+  [b] Frame as extension — builds on prior work, may seem incremental
+
+Select: a or b
+════════════════════════════════════════
+```
+
+### human-action Box
+
+```
+════════════════════════════════════════
+CHECKPOINT: human-action
+════════════════════════════════════════
+Task 4 of 6: Insert experimental results
+
+What is needed:
+  Exact p-value and effect size from your analysis
+
+Provide: p-value and effect size
+════════════════════════════════════════
+```
+
+## Status Symbols
+
+| Symbol | Meaning | Usage |
+|--------|---------|-------|
+| ✓ | Complete | Task finished, verification passed |
+| ► | In progress | Currently executing |
+| ✗ | Failed | Task or verification failed |
+| — | Skipped | Task skipped (gate disabled, not applicable) |
+| ⊘ | Blocked | Cannot proceed (missing dependency, auth) |
+
+### In Task Lists
+
+```
+  ✓ Task 1: Draft opening hook
+  ✓ Task 2: Establish research gap
+  ► Task 3: Present thesis statement
+  — Task 4: (skipped — gate disabled)
+  ⊘ Task 5: (blocked — needs Task 3)
+```
+
+## Progress Bars
+
+Format: `[{filled}{empty}] {current}/{total} {unit}`
+
+```
+[########............] 4/10 sections
+[####################] 10/10 sections
+[....................] 0/10 sections
+```
+
+Fill character: `#`
+Empty character: `.`
+Bar width: 20 characters (fixed)
+
+### In Section Tracking
+
+```
+Section progress:
+  [####................ ] 2/8 sections planned
+  [##.................. ] 1/8 sections written
+  [.................... ] 0/8 sections reviewed
+```
+
+### In Task Tracking
+
+```
+Plan 01-02 progress:
+  [############........] 3/5 tasks complete
+```
+
+## Color Note
+
+Claude Code renders plain text. All visual hierarchy comes from:
+
+- Unicode box-drawing characters (`━`, `═`, `─`) for borders
+- Unicode symbols (`✓`, `✗`, `►`, `⊘`, `—`) for status
+- Indentation and whitespace for structure
+- `#` and `.` for progress bars
+
+No ANSI color codes, no terminal escape sequences. These patterns work in any text rendering context.

package/core/write-the-f-paper/templates/UAT.md

@@ -0,0 +1,80 @@
+# UAT Template
+
+Template for `.planning/sections/XX-name/UAT.md` - section acceptance testing state file.
+
+---
+
+## File Template
+
+```markdown
+---
+status: testing
+section: XX-name
+started: YYYY-MM-DDTHH:MM:SSZ
+last_test: 0
+---
+
+## Current Test
+
+[Active test content will be displayed here during verification]
+
+## Tests
+
+### 1. [Criterion or claim text]
+result: pending
+reason: ""
+
+### 2. [Criterion or claim text]
+result: pending
+reason: ""
+
+[... one entry per test ...]
+
+## Summary
+
+total: N | passed: 0 | issues: 0 | skipped: 0
+
+## Gaps
+
+[Populated when tests result in "issue"]
+
+- claim: "Expected behavior or argument"
+  status: issue
+  reason: "User reported: ..."
+  severity: major
+```
+
+<guidelines>
+**Test sources:**
+
+Tests are derived from TWO sources:
+1. **PLAN.md `<success_criteria>` items** - Each checklist item becomes a test. These verify that the plan's deliverables were achieved.
+2. **argument-map.md claims mapped to this section** - Each claim assigned to this section becomes a test. These verify that the section fulfills its role in the paper's argument.
+
+**Each test is a single verifiable statement.** Do not combine multiple criteria into one test.
+
+**Status transitions:**
+- `pending` - Not yet evaluated
+- `pass` - User confirmed criterion is met
+- `skip` - User determined criterion is not applicable
+- `issue` - User identified a problem
+
+Transitions are one-way: `pending -> pass | skip | issue`. A test cannot return to pending.
+
+**Severity classification for gaps:**
+- `major` - Undermines the section's argument or misses a key claim. Must fix before submission.
+- `minor` - Weakens but does not invalidate. Should fix if time allows.
+- `cosmetic` - Polish issue (word choice, formatting, flow). Fix during final pass.
+
+**File location:** UAT.md lives in the section directory alongside PLAN.md and SUMMARY.md:
+```
+.planning/sections/XX-name/
+  XX-YY-PLAN.md
+  XX-YY-SUMMARY.md
+  UAT.md           <-- this file
+```
+
+**Persistence:** UAT.md survives `/clear`. On each invocation, verify-work reads UAT.md, finds the next `pending` test, presents it, records the result, and writes back. This enables verification to span multiple sessions.
+
+**Resumption:** The `last_test` field tracks the most recently evaluated test number. On resume, find the first test with `result: pending` (which may differ from last_test + 1 if tests were skipped non-sequentially).
+</guidelines>

package/core/write-the-f-paper/templates/base-prefs.yaml

@@ -0,0 +1,42 @@
+# WTF-P User Preferences
+# Global defaults — install to ~/.wtfp/base.yaml
+# Per-project overrides in .planning/prefs.yaml
+#
+# Inheritance: base.yaml → .planning/prefs.yaml → config.json (runtime)
+# Values here are defaults. Override per-project as needed.
+
+# Writing style preferences
+style:
+  voice: measured          # authoritative | measured | accessible | technical
+  person: third            # first | third | passive
+  hedging: moderate        # minimal | moderate | heavy
+  formality: academic      # casual | academic | formal
+
+# Citation habits
+citations:
+  style: apa               # apa | ieee | nature | chicago | vancouver
+  inline_format: author-year  # author-year | numbered | superscript
+  self_cite_limit: 3       # max self-citations before flagging
+
+# Prose preferences
+prose:
+  sentence_length: varied  # short | varied | long
+  paragraph_length: medium # short (3-5 sentences) | medium (5-8) | long (8-12)
+  jargon_level: moderate   # minimal | moderate | heavy
+  acronym_policy: define_first  # define_first | assume_known | footnote
+
+# Workflow defaults
+workflow:
+  confirm_before_write: true
+  auto_commit: true
+  backup_on_major_edit: true
+  research_before_plan: true
+
+# Model preferences (override config.json model_profile)
+# Uncomment to set global default
+# model_profile: balanced   # quality | balanced | budget
+
+# Editor integration
+editor:
+  preferred: none          # none | vscode | vim | emacs
+  open_after_write: false

package/core/write-the-f-paper/templates/config.json

@@ -3,28 +3,54 @@
   "depth": "standard",
   "document_type": "paper",
   "output_format": "markdown",
+  "model_profile": "balanced",
+  "venue_template": null,
   "gates": {
-    "confirm_project": true,
-    "confirm_sections": true,
-    "confirm_roadmap": true,
     "confirm_outline": true,
     "confirm_plan": true,
-    "write_next_plan": true,
+    "confirm_write": true,
+    "confirm_review": true,
+    "execute_next_section": true,
     "issues_review": true,
-    "confirm_transition": true
+    "confirm_transition": true,
+    "confirm_submission": true
   },
   "writing": {
     "claude_mode": "co-author",
     "citation_style": "apa",
     "verify_on_complete": true
   },
+  "workflow": {
+    "research": true,
+    "plan_check": true,
+    "verifier": true
+  },
   "verification": {
     "citation_check": true,
     "coherence_check": true,
     "rubric_check": true
   },
+  "planning": {
+    "commit_docs": true,
+    "search_gitignored": false
+  },
+  "parallelization": {
+    "enabled": false,
+    "plan_level": true,
+    "task_level": false,
+    "skip_checkpoints": true,
+    "max_concurrent_agents": 3,
+    "min_plans_for_parallel": 2
+  },
   "safety": {
     "always_confirm_destructive": true,
+    "always_confirm_external_services": true,
     "backup_before_major_edits": true
+  },
+  "git": {
+    "branching_strategy": "none",
+    "section_branch_template": "wtfp/section-{section}-{slug}",
+    "submission_branch_template": "wtfp/{milestone}-{slug}",
+    "squash_on_merge": false
   }
 }

package/core/write-the-f-paper/workflows/create-outline.wcn.md

@@ -142,7 +142,7 @@ WRITE: .planning/structure/narrative-arc.md
   USE: ~/.claude/write-the-f-paper/templates/project-context/narrative-arc.md
 [/step]
 
-[step:git_commit]
+[step:git_commit_initialization]
 RUN: git add .planning/PROJECT.md .planning/ROADMAP.md .planning/STATE.md .planning/sections/ .planning/structure/
 RUN: git add .planning/config.json 2>/dev/null
 COMMIT: "docs: initialize [title] ([N] sections)"