5-phase-workflow 1.9.5 → 2.0.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (44) hide show
  1. package/README.md +81 -410
  2. package/bin/install.js +385 -70
  3. package/bin/sync-agents.js +50 -11
  4. package/docs/findings.md +3 -3
  5. package/docs/workflow-guide.md +110 -1046
  6. package/package.json +6 -5
  7. package/src/agents/step-executor-agent.md +49 -0
  8. package/src/agents/step-orchestrator-agent.md +111 -0
  9. package/src/agents/verification-agent.md +78 -0
  10. package/src/commands/5/address-review-findings.md +69 -403
  11. package/src/commands/5/apply-review-findings.md +66 -0
  12. package/src/commands/5/configure.md +110 -76
  13. package/src/commands/5/discuss-feature.md +47 -57
  14. package/src/commands/5/eject.md +7 -6
  15. package/src/commands/5/implement.md +202 -0
  16. package/src/commands/5/plan.md +164 -0
  17. package/src/commands/5/reconfigure.md +32 -31
  18. package/src/commands/5/reply-pr-comments.md +46 -0
  19. package/src/commands/5/review.md +95 -0
  20. package/src/commands/5/split.md +190 -0
  21. package/src/commands/5/synchronize-agents.md +4 -4
  22. package/src/commands/5/triage-pr-comments.md +70 -0
  23. package/src/commands/5/update.md +8 -8
  24. package/src/hooks/check-updates.js +50 -7
  25. package/src/hooks/plan-guard.js +28 -22
  26. package/src/hooks/statusline.js +55 -4
  27. package/src/skills/configure-docs-index/SKILL.md +16 -21
  28. package/src/skills/configure-skills/SKILL.md +21 -24
  29. package/src/templates/AGENTS.md +94 -0
  30. package/src/templates/workflow/FIX-PLAN.md +1 -1
  31. package/src/templates/workflow/PLAN-COMPACT.md +42 -0
  32. package/src/templates/workflow/PLAN.md +58 -34
  33. package/src/templates/workflow/REVIEW-FINDINGS.md +7 -16
  34. package/src/templates/workflow/REVIEW-SUMMARY.md +5 -0
  35. package/src/templates/workflow/STATE.json +32 -3
  36. package/src/agents/component-executor.md +0 -57
  37. package/src/commands/5/implement-feature.md +0 -381
  38. package/src/commands/5/plan-feature.md +0 -293
  39. package/src/commands/5/plan-implementation.md +0 -333
  40. package/src/commands/5/quick-implement.md +0 -375
  41. package/src/commands/5/review-code.md +0 -212
  42. package/src/commands/5/verify-implementation.md +0 -277
  43. package/src/templates/workflow/FEATURE-SPEC.md +0 -100
  44. package/src/templates/workflow/VERIFICATION-REPORT.md +0 -103
package/README.md CHANGED
@@ -1,466 +1,137 @@
1
- # 5-Phase Workflow
1
+ # foifi
2
2
 
3
- A systematic, AI-assisted feature development workflow for Claude Code and Codex that works with any tech stack.
3
+ **foifi** is an opinionated AI development workflow layer that sits on top of Claude Code and Codex. It handles project setup, structured feature implementation, and code review — so you spend less time managing the AI and more time shipping.
4
4
 
5
- ## What is This?
5
+ ### What it does
6
6
 
7
- The **5-Phase Workflow** is a structured approach to feature development that breaks down the process into clear, manageable phases:
7
+ **Project setup** — The `/5:configure` command detects your stack, generates a `CLAUDE.md` / `AGENTS.md` tailored to your project, writes a `.5/index/` knowledge base, and installs project-specific skills and rules. This gives every AI session the right context from the start rather than letting the model guess.
8
8
 
9
- 1. **Feature Planning** - Understand requirements through intensive Q&A
10
- 2. **Implementation Planning** - Map requirements to technical components
11
- 3. **Orchestrated Implementation** - Execute with state tracking and parallel processing
12
- 4. **Verify Implementation** - Automated verification of completeness and correctness
13
- 5. **Code Review** - AI-powered review, findings annotation, and fix application
9
+ **Status line** foifi installs an informative Claude Code status line that surfaces the active feature, current workflow phase, and relevant state directly in the terminal footer. No more digging through files to remember where you left off.
14
10
 
15
- ## Why Use It?
11
+ **Structured implementation workflow** — Instead of asking Claude or Codex to "just implement this," foifi enforces a three-phase loop:
12
+ 1. **Plan** (`/5:plan`) — writes a single human-reviewed `plan.md` with scope, acceptance criteria, component checklist, and decisions.
13
+ 2. **Implement** (`/5:implement`) — an orchestrator agent turns the plan into a typed execution graph (`state.json`), then delegates each component to a focused executor agent. A verification agent checks completeness, correctness, and test coverage at the end of every run.
14
+ 3. **Review** (`/5:review`) — triages changed files, produces structured findings, and feeds them into `/5:address-review-findings` for interactive fix decisions and PR replies.
16
15
 
17
- - **Systematic**: Clear phases prevent missing requirements or skipping validation
18
- - **Efficient**: Parallel execution and smart agents minimize context usage
19
- - **Resumable**: State tracking allows pausing and continuing work across sessions
20
- - **Technology-Agnostic**: Works with JavaScript, Python, Java, Go, Rust, and more
21
- - **Transparent**: Visible progress tracking and clear handoffs between phases
16
+ This separation keeps planning readable, implementation mechanical, and review structured.
22
17
 
23
- ## Installation
18
+ **Code review and findings** — `/5:review` triages changed files and produces structured `review-findings-*.md`. `/5:address-review-findings` presents each finding interactively, records `fix`/`wont_fix`/`wait` decisions, applies approved local fixes, handles PR comment replies, and keeps a decision log — all without losing context between sessions.
24
19
 
25
- Install the workflow in your project using npx:
20
+ **Plan management helpers** — `/5:discuss-feature` refines an existing plan in conversation. `/5:split` breaks a large plan into smaller linked child plans. `/5:unlock` clears a stale planning lock. `/5:reconfigure` refreshes docs and skills when the project evolves.
26
21
 
27
- ```bash
28
- # Install locally for Claude Code
29
- npx 5-phase-workflow
30
-
31
- # Install locally for Codex
32
- npx 5-phase-workflow --codex
22
+ **Codex support** — Every command has a `$5-*` Codex equivalent. Codex runs are token-budgeted: simple steps use a lighter model and low reasoning, complex or security-sensitive steps escalate automatically.
33
23
 
34
- # Or install globally
35
- npx 5-phase-workflow --global
36
- npx 5-phase-workflow --codex --global
37
- ```
24
+ ### The name
38
25
 
39
- The installer will:
40
- - For Claude Code: copy workflow commands, agents, and skills to `.claude/`, then set up hooks and settings
41
- - For Codex: convert workflow commands into skills in `.codex/skills/` and generate `.codex/instructions.md`
42
- - Create `.5/features/` directory for feature tracking
26
+ "foifi" is Swiss German for *five*. The name comes from the project's original 5-phase workflow. That workflow has since been streamlined into the current 3-phase plan → implement → review loop, but the name stuck — and all commands still carry the `/5:` prefix.
43
27
 
44
- **After installation, you must configure your project:**
45
-
46
- ## Required: Configure Your Project
28
+ ## Install
47
29
 
48
30
  ```bash
49
- # Claude Code
50
- /5:configure
51
-
52
- # Codex
53
- $5-configure
31
+ npx foifi
32
+ npx foifi --codex
54
33
  ```
55
34
 
56
- This will:
57
- - Auto-detect your project type (JavaScript, Python, Java, etc.)
58
- - Set up build and test commands
59
- - Configure ticket tracking patterns
60
- - Generate comprehensive CLAUDE.md documentation
61
- - Generate a rebuildable codebase index in `.5/index/`
62
- - Create project-specific skills (create-component, create-service, etc.)
63
-
64
- Follow the standard workflow after configuration:
65
- 1. Claude Code: `/5:plan-implementation CONFIGURE`
66
- 2. Codex: `$5-plan-implementation CONFIGURE`
67
- 3. Claude Code: `/5:implement-feature CONFIGURE`
68
- 4. Codex: `$5-implement-feature CONFIGURE`
69
- 5. Claude Code: `/5:verify-implementation`
70
- 6. Codex: `$5-verify-implementation`
71
-
72
- **The workflow is ready to use after completing configuration.**
73
-
74
- ## Start Your First Feature
75
-
76
- After configuration is complete, start your first feature:
35
+ Global installs are also supported:
77
36
 
78
37
  ```bash
79
- # Claude Code
80
- /5:plan-feature
81
- /5:plan-implementation {ticket-id}-{description}
82
- /5:implement-feature {ticket-id}-{description}
83
- /5:verify-implementation
84
- /5:review-code
85
- /5:address-review-findings
86
-
87
- # Codex
88
- $5-plan-feature
89
- $5-plan-implementation {ticket-id}-{description}
90
- $5-implement-feature {ticket-id}-{description}
91
- $5-verify-implementation
92
- $5-review-code
93
- $5-address-review-findings
94
- ```
95
-
96
- **Tip:** Running `/clear` between phases in Claude Code resets context and keeps conversations focused. In Codex, start a fresh turn or keep the next phase focused. Each phase reads necessary artifacts from previous phases, so no context is lost.
97
-
98
- ## Supported Tech Stacks
99
-
100
- The workflow auto-detects and supports:
101
-
102
- **JavaScript/TypeScript**:
103
- - Node.js (npm, yarn, pnpm)
104
- - Next.js
105
- - NestJS
106
- - Express
107
- - React
108
- - Vue
109
-
110
- **Python**:
111
- - Django
112
- - Flask
113
- - FastAPI
114
- - Generic Python projects
115
-
116
- **Java**:
117
- - Gradle
118
- - Maven
119
- - Spring Boot
120
-
121
- **Other**:
122
- - Rust (Cargo)
123
- - Go
124
- - Ruby on Rails
125
- - Custom projects (via manual config)
126
-
127
- ## Available Commands
128
-
129
- Claude Code exposes the workflow under the `/5:` namespace. Codex exposes the same workflow as `$5-...` skills:
130
-
131
- | Command | Phase | Purpose |
132
- |---------|-------|---------|
133
- | `/5:configure` or `$5-configure` | Setup | Interactive project configuration |
134
- | `/5:plan-feature` or `$5-plan-feature` | 1 | Create feature specification with Q&A |
135
- | `/5:discuss-feature` or `$5-discuss-feature` | 1 | Refine existing feature spec |
136
- | `/5:plan-implementation` or `$5-plan-implementation` | 2 | Map feature to technical components |
137
- | `/5:implement-feature` or `$5-implement-feature` | 3 | Execute implementation with agents |
138
- | `/5:verify-implementation` or `$5-verify-implementation` | 4 | Verify completeness and correctness |
139
- | `/5:review-code` or `$5-review-code` | 5 | AI-powered code review (Claude, Codex, or CodeRabbit workflows) |
140
- | `/5:address-review-findings` or `$5-address-review-findings` | 5 | Apply annotated findings and address PR comments |
141
- | `/5:quick-implement` or `$5-quick-implement` | Fast | Streamlined workflow for small tasks |
142
- | `/5:eject` or `$5-eject` | Utility | Permanently remove update infrastructure |
143
- | `/5:unlock` or `$5-unlock` | Utility | Remove planning guard lock |
144
- | `/5:synchronize-agents` or `$5-synchronize-agents` | Utility | Sync user content between Claude Code and Codex runtimes |
145
-
146
- ## Configuration
147
-
148
- The workflow is configured via `.5/config.json`. Here's an example:
149
-
150
- ```json
151
- {
152
- "projectType": "nextjs",
153
- "ticket": {
154
- "pattern": "[A-Z]+-\\d+",
155
- "extractFromBranch": true
156
- },
157
- "build": {
158
- "command": "npm run build",
159
- "testCommand": "npm test"
160
- },
161
- "steps": [
162
- { "name": "foundation", "mode": "parallel" },
163
- { "name": "logic", "mode": "sequential" },
164
- { "name": "integration", "mode": "sequential" }
165
- ],
166
- "reviewTool": "coderabbit"
167
- }
38
+ npx foifi --global
39
+ npx foifi --codex --global
168
40
  ```
169
41
 
170
- ### Configuration Options
171
-
172
- #### Required Fields
173
-
174
- - **`projectType`**: Detected project type (e.g., `"nextjs"`, `"django"`, `"rust"`)
175
- - **`build.command`**: Command to build the project
176
- - **`build.testCommand`**: Command to run tests
177
-
178
- #### Optional Fields
179
-
180
- - **`ticket.pattern`**: Regex pattern for ticket IDs (e.g., `"PROJ-\\d+"`)
181
- - **`ticket.extractFromBranch`**: Auto-extract ticket from branch name
182
- - **`steps`**: Implementation step configuration
183
- - **`framework`**: Framework-specific patterns (routes, models, etc.)
184
- - **`integration`**: Integration point configuration
185
- - **`tools`**: Available development tools (CodeRabbit, IDE, etc.)
186
-
187
- Run `/5:configure` to set up or update your configuration.
188
-
189
- ## How It Works
190
-
191
- ### Phase 1: Feature Planning
192
-
193
- Claude asks 5-10 clarifying questions to understand your requirements:
194
-
195
- - What exactly should this feature do?
196
- - What are the edge cases?
197
- - How will we verify it works?
198
- - Are there simpler alternatives?
199
-
200
- The output is a comprehensive feature spec at `.5/features/{ticket-id}/feature.md`.
201
-
202
- ### Phase 2: Implementation Planning
203
-
204
- Claude maps your feature to technical components:
205
-
206
- - Analyzes your codebase structure
207
- - Identifies affected modules
208
- - Maps components to implementation steps
209
- - Creates dependency graph
210
-
211
- The output is an **atomic plan structure** at `.5/features/{ticket-id}/`:
212
- - `feature.md` - Feature specification (Phase 1)
213
- - `plan.md` - Implementation plan (Phase 2)
214
- - `state.json` - Implementation state tracking (Phase 3)
215
-
216
- Each step file is self-contained and independently loadable, making large plans manageable and improving agent efficiency.
217
-
218
- ### Phase 3: Orchestrated Implementation
219
-
220
- Claude executes the plan using specialized agents:
221
-
222
- - **step-executor**: Creates each component using skills or direct file creation
223
- - **step-verifier**: Compiles and checks for errors after each step
224
- - **integration-agent**: Wires components and registers routes
225
-
226
- State is tracked in `.5/features/{ticket-id}/state.json` for resumability.
227
-
228
- ### Phase 4: Verify Implementation
229
-
230
- An agent performs comprehensive verification:
231
-
232
- - All planned files exist
233
- - No compilation/build errors
234
- - Tests pass
235
- - IDE diagnostics clean
236
-
237
- Results are saved to a verification report.
238
-
239
- ### Phase 5: Code Review
240
-
241
- Two commands work together to handle the review workflow:
242
-
243
- **`/5:review-code`** — runs the automated review and presents findings:
244
- - Supports Claude (built-in, no setup) or CodeRabbit CLI
245
- - Reviews staged changes, unstaged changes, or branch diff
246
- - Categorizes findings as Fixable, Questions, or Manual
247
- - Lets you fix immediately or save findings for later
248
-
249
- **`/5:address-review-findings`** — applies annotated findings from a saved file:
250
- - Reads the `review-findings-*.md` file generated by `/5:review-code`
251
- - Applies `[FIX]` items, skips `[SKIP]` items, and follows `[MANUAL]` instructions
252
- - Optionally fetches and addresses GitHub PR review comments
253
- - Posts threaded replies to processed PR comments
254
- - Runs build, tests, and lint after applying fixes
255
- - Saves a summary report
42
+ Claude Code installs commands under `.claude/`. Codex installs converted skills under `.codex/skills/`.
256
43
 
257
- ## Project Structure
44
+ ## Configure
258
45
 
259
- After installation, your `.claude/` directory will contain:
260
-
261
- ```
262
- .5/
263
- ├── config.json # Project configuration
264
- ├── version.json # Version tracking
265
- ├── index/ # Generated codebase index + rebuild script
266
- │ ├── rebuild-index.sh
267
- │ └── *.md
268
- └── features/ # Feature tracking
269
-
270
- .claude/
271
- ├── commands/5/ # Workflow commands
272
- │ ├── plan-feature.md
273
- │ ├── plan-implementation.md
274
- │ ├── implement-feature.md
275
- │ ├── verify-implementation.md
276
- │ ├── review-code.md
277
- │ ├── address-review-findings.md
278
- │ ├── discuss-feature.md
279
- │ ├── quick-implement.md
280
- │ ├── configure.md
281
- │ ├── eject.md
282
- │ ├── unlock.md
283
- │ └── synchronize-agents.md
284
- ├── skills/ # Atomic operations
285
- │ ├── build-project/
286
- │ ├── run-tests/
287
- │ └── generate-readme/
288
- ├── hooks/
289
- │ ├── statusline.js # Status line integration
290
- │ ├── check-updates.js # Update notifications
291
- │ ├── plan-guard.js # Planning phase edit guard
292
- │ └── config-guard.js # Configuration guard
293
- └── settings.json # Claude Code settings
294
- ```
295
-
296
- ## Examples
297
-
298
- ### Example 1: Adding a REST API endpoint
46
+ Run configuration once per project:
299
47
 
300
48
  ```bash
301
- /5:plan-feature
302
- # Claude asks about the endpoint: path, methods, request/response format, validation, etc.
303
- # Creates feature spec
304
-
305
- /5:plan-implementation PROJ-1234-add-user-endpoint
306
- # Claude maps to: route file, controller, service, tests
307
- # Creates technical plan
308
-
309
- /5:implement-feature PROJ-1234-add-user-endpoint
310
- # Creates route, controller, service, tests
311
- # Registers route in app
312
- # Runs build and tests
313
-
314
- /5:verify-implementation
315
- # Verifies all files created
316
- # Checks build passes
317
- # Confirms tests pass
49
+ /5:configure
50
+ # or in Codex
51
+ $5-configure
318
52
  ```
319
53
 
320
- ### Example 2: Small bug fix
54
+ Configuration writes `.5/config.json` and `.5/features/CONFIGURE/plan.md`. Then run:
321
55
 
322
56
  ```bash
323
- /5:quick-implement
324
- # Describe the fix
325
- # Claude implements, builds, tests in one step
57
+ /5:implement CONFIGURE
58
+ # or
59
+ $5-implement CONFIGURE
326
60
  ```
327
61
 
328
- ## Troubleshooting
329
-
330
- ### Build/Test commands not working
331
-
332
- 1. Run `/5:configure` to verify configuration
333
- 2. Test commands manually in terminal
334
- 3. Update `.5/config.json` with correct commands
335
-
336
- ### "Cannot find project type"
62
+ That generates project documentation, a rebuildable `.5/index/`, AGENTS.md, a CLAUDE.md shim, and selected project-specific skills/rules.
337
63
 
338
- The auto-detection failed. Run `/5:configure` and manually select your project type.
339
-
340
- ### State file issues
341
-
342
- If implementation gets stuck:
343
-
344
- 1. Check `.5/features/{ticket-id}/state.json`
345
- 2. Note the `currentStep` value
346
- 3. Run `/5:implement-feature` again - it will resume from that step
347
-
348
- ### CodeRabbit not working
349
-
350
- 1. Install: https://docs.coderabbit.ai/cli/installation
351
- 2. Authenticate: `coderabbit auth login`
352
- 3. Run `/5:configure` to update config
353
-
354
- ## Updating
355
-
356
- The workflow automatically detects when a new version is available.
357
-
358
- ### Automatic Update (Recommended)
64
+ ## Workflow
359
65
 
360
66
  ```bash
361
- # Interactive upgrade (shows prompt)
362
- npx 5-phase-workflow
363
-
364
- # Force upgrade (no prompts)
365
- npx 5-phase-workflow --upgrade
366
-
367
- # Check version without updating
368
- npx 5-phase-workflow --check
67
+ /5:plan
68
+ /5:split {feature-name}
69
+ /5:implement {feature-name}
70
+ /5:review
71
+ /5:address-review-findings {feature-name}
369
72
  ```
370
73
 
371
- ### Legacy Upgrade Method
74
+ Codex equivalents:
372
75
 
373
76
  ```bash
374
- npx 5-phase-workflow --uninstall
375
- npx 5-phase-workflow
77
+ $5-plan
78
+ $5-split {feature-name}
79
+ $5-implement {feature-name}
80
+ $5-review
81
+ $5-address-review-findings {feature-name}
376
82
  ```
377
83
 
378
- **Note:** During updates:
379
- - Config files in `.5/` are preserved
380
- - User-created commands, agents, skills, hooks, and templates are preserved
381
- - Only workflow-managed files are updated
382
-
383
- ### Ejecting
84
+ Verification runs at the end of `/5:implement` and records concise results in `state.json`.
384
85
 
385
- If you want to permanently opt out of the update system (e.g., to customize workflow files without future updates overwriting them), run:
86
+ ## Commands
386
87
 
387
- ```bash
388
- # Claude Code
389
- /5:eject
390
-
391
- # Codex
392
- $5-eject
393
- ```
88
+ | Command | Purpose |
89
+ |---------|---------|
90
+ | `/5:configure` / `$5-configure` | Detect project settings and write the CONFIGURE plan |
91
+ | `/5:plan` / `$5-plan` | Create one unified `plan.md` from requirements, codebase exploration, and user decisions |
92
+ | `/5:discuss-feature` / `$5-discuss-feature` | Refine an existing `plan.md` |
93
+ | `/5:split` / `$5-split` | Split an existing plan into smaller linked plans for separate implementation |
94
+ | `/5:implement` / `$5-implement` | Derive `state.json`, execute steps with agents, and verify inline |
95
+ | `/5:review` / `$5-review` | Review code changes and save findings |
96
+ | `/5:address-review-findings` / `$5-address-review-findings` | Decide on review findings interactively, then apply approved fixes and PR comments |
97
+ | `/5:reconfigure` / `$5-reconfigure` | Refresh docs, index, skills, and rules |
98
+ | `/5:update` / `$5-update` | Upgrade installed workflow files |
99
+ | `/5:eject` / `$5-eject` | Stop workflow-managed updates |
100
+ | `/5:unlock` / `$5-unlock` | Remove a stale planning guard lock |
101
+ | `/5:synchronize-agents` / `$5-synchronize-agents` | Sync user content between Claude Code and Codex |
394
102
 
395
- This permanently removes the update infrastructure:
396
- - Deletes `check-updates.js` hook, `update.md` and `eject.md` commands
397
- - Deletes `.5/version.json` and `.5/.update-cache.json`
398
- - For Claude Code, removes the update check hook entry from `.claude/settings.json`
399
- - For Codex, removes the converted update/eject skills from `.codex/skills/`
103
+ ## Artifacts
400
104
 
401
- All other workflow files remain untouched. **This is irreversible.** To restore update functionality, reinstall with `npx 5-phase-workflow`.
105
+ Each feature lives under `.5/features/{feature-name}/`:
402
106
 
403
- ### Synchronizing Runtimes
107
+ - `plan.md` - single human-reviewed planning artifact
108
+ - `codebase-scan.md` - cached discovery used to reduce repeated scanning
109
+ - `state.json` - enriched execution state derived by `step-orchestrator-agent`
110
+ - `state-events.jsonl` - detailed execution history for retries, commands, commits, and verification
111
+ - `split-manifest-*.json` - parent feature record for child plans created by `/5:split`
112
+ - `review-findings-*.md` - review output for `/5:address-review-findings`
113
+ - `review-decisions-*.json` - interactive fix/wont-fix/wait decisions for local findings
114
+ - `pr-comment-decisions.json` - PR review comment decisions when PR handling is used
404
115
 
405
- If you have both Claude Code and Codex installed, user-generated content (project-specific skills, custom commands, rules) only exists in the runtime where it was created. To sync this content bidirectionally:
116
+ ## Design
406
117
 
407
- ```bash
408
- # Claude Code
409
- /5:synchronize-agents
118
+ Planning stays human-readable. `plan.md` contains scope, acceptance criteria, decisions, module impact, and a clean component checklist. Small low-risk changes can use the compact plan template. Plans intentionally do not ask the planner to fill model choices, verify commands, step grouping, or pattern-file wiring.
410
119
 
411
- # Codex
412
- $5-synchronize-agents
413
- ```
120
+ Implementation is mechanical. `step-orchestrator-agent` reads `plan.md` and `codebase-scan.md`, derives the execution graph into compact `state.json`, then `/5:implement` delegates each component to `step-executor-agent` with an inline executor contract. Pattern context is passed as targeted references instead of broad file lists so executors can read only the relevant ranges. Detailed history is appended to `state-events.jsonl`. This reduces planning token cost and avoids brittle prompt-table metadata.
414
121
 
415
- This will:
416
- - Sync project-specific skills (e.g., `create-controller`, `run-tests`) between `.claude/skills/` and `.codex/skills/` with appropriate format conversion
417
- - Convert custom Claude commands to Codex skills
418
- - Append `.claude/rules/` content to `.codex/instructions.md`
419
- - Sync Codex-only skills back to Claude
122
+ Verification uses a dedicated agent. `/5:implement` runs `verification-agent` at the end and records a concise final status in `state.json` without generating an extra report.
420
123
 
421
- Workflow-managed files and shared data (`.5/`) are not affected those are handled by the installer.
124
+ Review is risk-based. Native review triages changed files first and reads full files only for risky changes or when diff context is insufficient. `/5:address-review-findings` presents each finding one by one with a recommendation, records `fix`/`wont_fix`/`wait` decisions, then coordinates narrower helpers for approved local fixes, PR comment triage, and PR replies so the common path stays compact.
422
125
 
423
- ## Development
126
+ Reconfiguration uses a compact `.5/reconfigure-manifest.json` to pass refresh decisions to documentation and skill generation helpers without duplicating long detection summaries in prompts.
424
127
 
425
- ### Running Tests
128
+ For Codex installs, the workflow is token-budgeted: exploration, orchestration, and simple executors default to `gpt-5.4-mini` with low reasoning. Complex logic, security-sensitive work, data migrations, public API changes, final verification that needs deeper review, and failed retries escalate to `gpt-5.4` with medium reasoning.
426
129
 
427
- The project includes automated verification to ensure all workflow files are properly configured:
130
+ ## Updating
428
131
 
429
132
  ```bash
430
- # Run verification tests
431
- npm test
432
-
433
- # Or run directly
434
- bash test/verify-install-js.sh
133
+ npx foifi --upgrade
134
+ npx foifi --codex --upgrade
435
135
  ```
436
136
 
437
- This verifies that all workflow files (commands, agents, skills, hooks, templates) are properly listed in `bin/install.js` for selective updates.
438
-
439
- ### Continuous Integration
440
-
441
- A GitHub Actions workflow runs on every push to verify the install.js configuration. The workflow:
442
- - Checks that all workflow files are listed in `getWorkflowManagedFiles()`
443
- - Ensures selective updates will work correctly
444
- - Prevents accidental omissions that could break user upgrades
445
-
446
- See `.github/workflows/test.yml` for details.
447
-
448
- ### Adding New Workflow Files
449
-
450
- When adding new commands, agents, skills, hooks, or templates:
451
-
452
- 1. Create the file in the appropriate `src/` directory
453
- 2. **Update `bin/install.js`** - Add the file to `getWorkflowManagedFiles()`
454
- 3. Run `npm test` to verify
455
- 4. Commit only if tests pass
456
-
457
- See `CLAUDE.md` for detailed development guidelines.
458
-
459
- ## License
460
-
461
- MIT
462
-
463
- ## Learn More
464
-
465
- - [Full Workflow Guide](./docs/workflow-guide.md) - Detailed documentation
466
- - [Claude Code Docs](https://docs.anthropic.com/claude/docs/claude-code) - Claude Code features
137
+ v2.0.0 is a hard migration. Finish in-progress v1.9.5 features before upgrading; v1 `feature.md` and old `state.json` formats are not supported.