create-harness-vibe-coding 0.8.8 → 0.8.10

package/templates/common/README.md

@@ -37,7 +37,7 @@ The agentic engineering harness lives in `Harness/`.
 - Start at `Harness/README.md`.
 - Load memory and resource registrations from `Harness/MEMORY.md`.
 - Track active work in `Harness/PROGRESS.md` and `Harness/tasks/<task-id>/PROGRESS.md`.
-- Use `Harness/WF.md` for long, difficult, multi-agent work.
+- Use `Harness/WF.md` only when the user explicitly invokes a WF command such as `/wf` or `/wf-max`; complex work may still use direct planning, tests, and subagents without entering WF.
   - Claude Code: invoke the `wf` skill with `/wf`.
   - Codex: invoke the `wf` skill with `$wf` or `/skills`.
 - Use `Harness/subagents.md` when coordinating multiple agents.
@@ -45,4 +45,4 @@ The agentic engineering harness lives in `Harness/`.
 Tool discovery files stay at the repository root:
 
 - Claude Code: `.claude/settings.json`, `.claude/agents/`, and `.claude/skills/`.
-- Codex: `.agents/skills/` for repo skills and `.codex/` for config placeholders. Runtime hooks are absent by default; only `/wf-auto` may opt into a bounded tick hook.
+- Codex: `.agents/skills/` for repo skills and `.codex/` for config placeholders. The bundled update reminder uses a startup-only hook; avoid turn-by-turn runtime hooks unless `/wf-auto` explicitly opts into a bounded tick hook.

package/templates/common/memory/agent-lessons-patterns.md

@@ -7,15 +7,16 @@ Write here when:
 - A validation failure exposes a missing regression check.
 - A handoff, dispatch, or context-loading pattern should be repeated or avoided.
 
-Entry format, newest first:
+Entry format (compact, default no date):
 
 ```markdown
-## YYYY-MM-DD - Short Lesson Name
-
-- Lesson: the reusable pattern.
-- Source: review finding, debug loop, failed verification, or handoff.
-- Apply when: the task shape or files where this matters.
-- Regression guard: test, validator check, docs update, or manual evidence to keep it from recurring.
+- When <scenario>: <rule>. Avoid <over-application>. Signals: <signals>.
 ```
 
-Keep entries lightweight and actionable. Avoid secrets and speculative lessons.
+Only use date/timestamp headings when:
+- Entry supersedes prior conflicting guidance
+- Time-sensitive context (version, deprecation)
+- Conflict resolution needed
+
+Keep entries lightweight and actionable. Avoid secrets, speculative lessons, task logs, and process summaries.
+- Entry supersedes prior conflicting guidance: add date stamp.

package/templates/common/memory/routes.md

@@ -0,0 +1,43 @@
+# Memory Routes (L3 Route Index)
+
+> L3 = 详细 durable memory 路由索引。不是详细 memory，而是命中规则索引。
+> 详细 memory 文件按 scenario 匹配后按需加载。
+
+| id | signals | load | avoid |
+| --- | --- | --- | --- |
+| explicit-user-preference | remember, next time, 下次, 记住, 不要再, never, always, I prefer | user-corrections-preferences.md | current-task-only notes, transient mood |
+| hooks-only-wf-auto | hook, runtime hook, .claude/settings.json, .codex/hooks.json, WF-AUTO.md | agent-lessons-patterns.md#only-wf-auto-may-use-runtime-hooks | non-runtime docs, broad hook designs |
+| tool-repeat-failure | same command fails 3x, error signature repeats | tool-usage-reflections.md | one-off command failures |
+| review-debug-lesson | reusable review/debug/validation pattern | agent-lessons-patterns.md | task-specific logs |
+| startup-digest | new session, startup | startup-hints.md | full MEMORY.md index |
+| wf-closeout | /wf-learn, workflow closeout | all memory files via context-master -> memory-master | premature writes before extraction |
+
+## Matching Algorithm
+
+1. 把当前请求压缩成 **scenario pack**: intent / files / commands / signals / risk
+2. 与上表 Signals 列匹配，计算 score:
+   - +4 exact file/path match
+   - +3 explicit trigger phrase match
+   - +3 command/error signature match
+   - +2 workflow/mode match
+   - +1 keyword overlap
+   - -4 avoid match
+3. 阈值:
+   - score >= 5: 加载对应 memory entry/file section
+   - score 3-4: 只加载 route 摘要，必要时再读详细文件
+   - score < 3: 不加载
+4. Hard trigger 优先:
+   - 用户显式记忆语句 -> user-corrections-preferences.md
+   - 同命令失败 3 次 -> tool-usage-reflections.md
+   - review/debug/validation 可复用 lesson -> agent-lessons-patterns.md
+   - 修改 hook 相关文件 -> 加载 only-wf-auto hook lesson
+   - /wf-learn -> 完整 learning cycle
+
+## Load Logging
+
+每次加载 memory hints 需说明原因:
+```
+memory hints loaded: <id> because <signal>
+```
+
+不使用 embedding；保持稳定、可解释、省 token。

package/templates/common/memory/startup-hints.md

@@ -0,0 +1,32 @@
+# Startup Hints (L2 Memory Digest)
+
+> L2 = 启动极简 memory digest。每次 Harness 会话启动可读此文件，但不要加载完整 Harness/MEMORY.md 或 Harness/README.md。
+
+## Core Principles
+
+- Memory 是短场景规则，不是 task log。
+- 时间戳默认不要写；只有 superseded / conflict / time-sensitive 才写。
+- Broad runtime hooks 禁止；只有 `/wf-auto` bounded tick hook 例外。
+- Direct mode 不加载完整 Harness router，但 startup-hints 是允许的轻量启动提示。
+- 详细 memory 仍按 `MEMORY_PROTOCOL.md` 场景命中后再读。
+
+## Memory Candidate Detection
+
+用户说以下触发词时，识别为 memory candidate：
+
+**英文**: remember, next time, don't, do not, never, always, I prefer, I want you to
+**中文**: 记住, 下次, 以后, 不要再, 总是, 永远不要, 我偏好, 我希望你以后
+
+## When to Write Memory
+
+- **Explicit user preference**: 清晰、安全、场景明确时，可立即写入 L3，不必等 `/wf-learn`。
+- **Repeated implicit correction**: 同一假设/模式被纠正 2+ 次。
+- **Tool/command failure**: 同类工具/命令失败 3+ 次。
+- **Review/debug lesson**: 只有可复用、能防回归时才写。
+
+## What NOT to Write
+
+- Task log、过程总结、一次性情绪
+- Raw logs、transcripts
+- Secrets, tokens, credentials, private data
+- 临时偏好、无复用价值的笔记

package/templates/common/memory/tool-usage-reflections.md

@@ -7,15 +7,16 @@ Write here when:
 - A more reliable command pattern replaces a brittle one.
 - The environment needs a durable fix, flag, path rule, shell syntax, or startup sequence.
 
-Entry format, newest first:
+Entry format (compact, default no date):
 
 ```markdown
-## YYYY-MM-DD - Short Pattern Name
-
-- Trigger: what failed or repeated.
-- Better pattern: the command, tool usage, or sequence to use next time.
-- Evidence: command output summary, error text, or affected environment.
-- Scope: when this applies and when it does not.
+- When <scenario>: <rule>. Avoid <over-application>. Signals: <signals>.
 ```
 
-Keep entries concise. Do not record secrets, credentials, private tokens, or one-off noise.
+Only use date/timestamp headings when:
+- Entry supersedes prior conflicting guidance
+- Time-sensitive context (version, deprecation)
+- Conflict resolution needed
+
+Never record one-off command failures. Never store secrets, credentials, or private tokens.
+- Entry supersedes prior conflicting guidance: add date stamp.

package/templates/common/memory/user-corrections-preferences.md

@@ -3,19 +3,21 @@
 Purpose: record repeated user corrections, durable preferences, and common-sense course corrections.
 
 Write here when:
-- The user says "remember", "never", "next time", "always", or "I prefer".
+- The user says "remember", "never", "next time", "always", "I prefer", "记住", "下次", "不要再" etc.
+- Explicit, safe, scoped user preferences can be written immediately without `/wf-learn`.
 - The user corrects the same assumption/pattern 2+ times.
 - A correction changes how future work should be scoped, explained, verified, or handed off.
 
-Entry format, newest first:
+Entry format (compact, default no date):
 
 ```markdown
-## YYYY-MM-DD - Short Preference Name
-
-- Correction/preference: the durable instruction.
-- Trigger: what prompted the correction.
-- Apply when: future contexts where this should guide behavior.
-- Avoid: contexts where this should not be over-applied.
+- When <scenario>: <rule>. Avoid <over-application>. Signals: <signals>.
 ```
 
-Do not record ordinary chat. If the preference is ambiguous, ask before writing it. Never store secrets.
+Only use date/timestamp headings when:
+- Entry supersedes prior conflicting guidance
+- Time-sensitive context (version, deprecation)
+- Conflict resolution needed
+
+Do not record ordinary chat, task logs, process summaries, one-time emotions, or transient preferences. If the preference is ambiguous, ask before writing it. Never store secrets.
+- Entry supersedes prior conflicting guidance: add date stamp.

package/templates/optional/catalog.json

@@ -73,6 +73,14 @@
       "url": "https://github.com/colbymchenry/codegraph",
       "tags": ["code graph", "architecture", "analysis"]
     },
+    {
+      "id": "grill-me",
+      "title": "Grill Me",
+      "description": "External Matt Pocock skill for relentlessly interviewing a plan or design before implementation.",
+      "installMode": "recommend-only",
+      "url": "https://github.com/mattpocock/skills/tree/main/skills/productivity/grill-me",
+      "tags": ["planning", "design", "interview", "skills"]
+    },
     {
       "id": "oh-my-openagent",
       "title": "oh-my-openagent",

package/templates/optional/skills/browser-e2e/.claude/skills/browser-e2e/SKILL.md

@@ -1,42 +1,42 @@
----
-name: browser-e2e
-description: Mandatory real-browser smoke and end-to-end verification using Playwright, Chrome DevTools/CDP, or documented manual browser evidence.
----
-
-# Browser E2E
-
-## When To Use
-
-Use this skill when a change affects browser-visible behavior, navigation, forms, routing, layout, or client-side integration. Web/UI acceptance requires loading the app in a real browser before claiming the UI is done.
-
-## Docs To Load
-
-- `Harness/workflows/browser-e2e.md`
-- `Harness/PROGRESS.md` and current task `tasks/<id>/PLAN.md`
-- Existing project test, build, and run instructions.
-
-## Required Inputs
-
-- Target URL or command to start the app.
-- User flows or pages to verify.
-- Expected behavior and supported viewport/browser scope.
-- Selector contract: stable accessible labels/roles and `data-testid` hooks for critical inputs, buttons, filters, rows, empty/error/loading states, and other targetable UI states.
-
-## Allowed Writes
-
-- Browser test files in the project's existing test locations.
-- Evidence artifacts such as screenshots, traces, or reports in existing artifact folders.
-- Notes in current task plan when the active task asks for plan tracking.
-
-## Output Format
-
-Return changed files, commands run, browser evidence paths, verified flows, failures, and follow-up risks.
-Include the selectors used for CDP/Playwright/manual verification.
-
-## PROGRESS.md & Task PLAN.md Updates
-
-Update `Harness/tasks/<task-id>/PLAN.md` only when executing a tracked plan item or recording evidence requested by the current task.
-
-## dispatch.md Usage
-
-Use `Harness/dispatch.md` only when splitting independent browser checks across workers is explicitly useful.
+---
+name: browser-e2e
+description: Mandatory real-browser smoke and end-to-end verification using Playwright, Chrome DevTools/CDP, or documented manual browser evidence.
+---
+
+# Browser E2E
+
+## When To Use
+
+Use this skill when a change affects browser-visible behavior, navigation, forms, routing, layout, or client-side integration. Web/UI acceptance requires loading the app in a real browser before claiming the UI is done.
+
+## Docs To Load
+
+- `Harness/workflows/browser-e2e.md`
+- `Harness/PROGRESS.md` and current task `tasks/<id>/PLAN.md`
+- Existing project test, build, and run instructions.
+
+## Required Inputs
+
+- Target URL or command to start the app.
+- User flows or pages to verify.
+- Expected behavior and supported viewport/browser scope.
+- Selector contract: stable accessible labels/roles and `data-testid` hooks for critical inputs, buttons, filters, rows, empty/error/loading states, and other targetable UI states.
+
+## Allowed Writes
+
+- Browser test files in the project's existing test locations.
+- Evidence artifacts such as screenshots, traces, or reports in existing artifact folders.
+- Notes in current task plan when the active task asks for plan tracking.
+
+## Output Format
+
+Return changed files, commands run, browser evidence paths, verified flows, failures, and follow-up risks.
+Include the selectors used for CDP/Playwright/manual verification.
+
+## PROGRESS.md & Task PLAN.md Updates
+
+Update `Harness/tasks/<task-id>/PLAN.md` only when executing a tracked plan item or recording evidence requested by the current task.
+
+## dispatch.md Usage
+
+Use `Harness/dispatch.md` only when splitting independent browser checks across workers is explicitly useful.

package/templates/optional/skills/browser-e2e/.opencode/commands/wf-browser.md

@@ -0,0 +1,15 @@
+---
+description: Run AI-driven browser automation and E2E testing via the wf-browser skill
+---
+
+# /wf-browser
+
+This is a **workflow command**, not a direct command. Do not execute it as a
+static help or script command.
+
+1. Load `CLAUDE.md`, `Harness/MEMORY.md` (index only per Memory Preflight), then `Harness/README.md`.
+2. Execute per the skill adapter `.claude/skills/wf-browser/SKILL.md` (mirror: `.agents/skills/wf-browser/SKILL.md`).
+3. Do not duplicate the workflow here. The skill adapter and `Harness/workflows/browser-e2e.md` are authoritative.
+
+If this runtime cannot invoke the skill directly, read
+`.claude/skills/wf-browser/SKILL.md` and follow it in place.

package/templates/optional/skills/github-pr-review/.claude/skills/github-pr-review/SKILL.md

@@ -1,40 +1,40 @@
----
-name: github-pr-review
-description: GitHub pull request review workflow using available GitHub CLI, local git diff, checks, and CI evidence.
----
-
-# GitHub PR Review
-
-## When To Use
-
-Use this skill when reviewing a GitHub pull request, responding to PR feedback, checking CI status, or summarizing review findings.
-
-## Docs To Load
-
-- `Harness/workflows/github-pr-review.md`
-- `Harness/PROGRESS.md` and current task `tasks/<id>/PLAN.md`
-- Repository contribution, test, and review guidelines.
-
-## Required Inputs
-
-- PR number, branch, or comparison range.
-- Review goal: bug hunt, approval readiness, CI diagnosis, or feedback response.
-- Expected test and check requirements.
-
-## Allowed Writes
-
-- Local files needed to address approved review feedback.
-- Review notes or evidence in existing docs only when requested.
-- Current task `tasks/<id>/PLAN.md` when tracking review work.
-
-## Output Format
-
-Return findings first by severity with file and line references, then open questions, tests/checks run, and change summary if edits were made.
-
-## PROGRESS.md & Task PLAN.md Updates
-
-Update `Harness/tasks/<task-id>/PLAN.md` only when the PR review is part of a tracked implementation plan.
-
-## dispatch.md Usage
-
-Use `Harness/dispatch.md` when independent review areas can be assigned separately, such as backend, frontend, and CI.
+---
+name: github-pr-review
+description: GitHub pull request review workflow using available GitHub CLI, local git diff, checks, and CI evidence.
+---
+
+# GitHub PR Review
+
+## When To Use
+
+Use this skill when reviewing a GitHub pull request, responding to PR feedback, checking CI status, or summarizing review findings.
+
+## Docs To Load
+
+- `Harness/workflows/github-pr-review.md`
+- `Harness/PROGRESS.md` and current task `tasks/<id>/PLAN.md`
+- Repository contribution, test, and review guidelines.
+
+## Required Inputs
+
+- PR number, branch, or comparison range.
+- Review goal: bug hunt, approval readiness, CI diagnosis, or feedback response.
+- Expected test and check requirements.
+
+## Allowed Writes
+
+- Local files needed to address approved review feedback.
+- Review notes or evidence in existing docs only when requested.
+- Current task `tasks/<id>/PLAN.md` when tracking review work.
+
+## Output Format
+
+Return findings first by severity with file and line references, then open questions, tests/checks run, and change summary if edits were made.
+
+## PROGRESS.md & Task PLAN.md Updates
+
+Update `Harness/tasks/<task-id>/PLAN.md` only when the PR review is part of a tracked implementation plan.
+
+## dispatch.md Usage
+
+Use `Harness/dispatch.md` when independent review areas can be assigned separately, such as backend, frontend, and CI.

package/templates/optional/skills/python-backend/.claude/skills/python-backend/SKILL.md

@@ -1,40 +1,40 @@
----
-name: python-backend
-description: Python backend workflow for API changes, service logic, FastAPI-style routes, and unittest or pytest verification.
----
-
-# Python Backend
-
-## When To Use
-
-Use this skill when changing Python API routes, service logic, persistence boundaries, background jobs, or backend tests.
-
-## Docs To Load
-
-- `Harness/workflows/python-backend.md`
-- `Harness/PROGRESS.md` and current task `tasks/<id>/PLAN.md`
-- Project backend README, API docs, dependency files, and test configuration.
-
-## Required Inputs
-
-- Target behavior, endpoint, service, or bug.
-- Existing test command and runtime setup.
-- Database, environment variable, or fixture constraints.
-
-## Allowed Writes
-
-- Python source, tests, fixtures, and backend docs in the task scope.
-- Local evidence artifacts such as test output snippets.
-- Current task `tasks/<id>/PLAN.md` when the task requires plan tracking.
-
-## Output Format
-
-Return files changed, API or behavior summary, tests run, important logs, migration/config notes, and remaining risks.
-
-## PROGRESS.md & Task PLAN.md Updates
-
-Update `Harness/tasks/<task-id>/PLAN.md` only when executing a tracked plan item or recording required validation evidence.
-
-## dispatch.md Usage
-
-Use `Harness/dispatch.md` for separable backend work such as API implementation, database changes, and test coverage.
+---
+name: python-backend
+description: Python backend workflow for API changes, service logic, FastAPI-style routes, and unittest or pytest verification.
+---
+
+# Python Backend
+
+## When To Use
+
+Use this skill when changing Python API routes, service logic, persistence boundaries, background jobs, or backend tests.
+
+## Docs To Load
+
+- `Harness/workflows/python-backend.md`
+- `Harness/PROGRESS.md` and current task `tasks/<id>/PLAN.md`
+- Project backend README, API docs, dependency files, and test configuration.
+
+## Required Inputs
+
+- Target behavior, endpoint, service, or bug.
+- Existing test command and runtime setup.
+- Database, environment variable, or fixture constraints.
+
+## Allowed Writes
+
+- Python source, tests, fixtures, and backend docs in the task scope.
+- Local evidence artifacts such as test output snippets.
+- Current task `tasks/<id>/PLAN.md` when the task requires plan tracking.
+
+## Output Format
+
+Return files changed, API or behavior summary, tests run, important logs, migration/config notes, and remaining risks.
+
+## PROGRESS.md & Task PLAN.md Updates
+
+Update `Harness/tasks/<task-id>/PLAN.md` only when executing a tracked plan item or recording required validation evidence.
+
+## dispatch.md Usage
+
+Use `Harness/dispatch.md` for separable backend work such as API implementation, database changes, and test coverage.

package/templates/optional/skills/ts-react-frontend/.claude/skills/ts-react-frontend/SKILL.md

@@ -1,43 +1,43 @@
----
-name: ts-react-frontend
-description: TypeScript React frontend workflow for components, state, routing, typecheck, tests, builds, and browser smoke evidence.
----
-
-# TypeScript React Frontend
-
-## When To Use
-
-Use this skill when changing React components, hooks, routes, client state, styling, build configuration, or frontend tests.
-For user-visible changes, typecheck, build, and unit tests are not enough; include real-browser smoke or screenshot evidence before claiming acceptance.
-
-## Docs To Load
-
-- `Harness/workflows/ts-react-frontend.md`
-- `Harness/PROGRESS.md` and current task `tasks/<id>/PLAN.md`
-- Existing frontend README, package scripts, design system, and test setup.
-
-## Required Inputs
-
-- Screen, component, or flow being changed.
-- Expected behavior and acceptance criteria.
-- Existing package manager and verification commands.
-- Selector contract: stable accessible labels/roles and `data-testid` hooks for critical inputs, buttons, filters, rows, empty/error/loading states, and other targetable UI states.
-
-## Allowed Writes
-
-- Frontend source, styles, tests, and focused docs in task scope.
-- Generated evidence such as screenshots or test reports in existing artifact paths.
-- Current task `tasks/<id>/PLAN.md` when tracking a plan item.
-
-## Output Format
-
-Return changed files, UI behavior summary, commands run, browser or test evidence, and unresolved risks.
-Include selectors added or verified for CDP/Playwright/manual checks.
-
-## PROGRESS.md & Task PLAN.md Updates
-
-Update `Harness/tasks/<task-id>/PLAN.md` only when executing a tracked task or recording required validation evidence.
-
-## dispatch.md Usage
-
-Use `Harness/dispatch.md` when independent frontend tasks can run in parallel, such as components, tests, and browser checks.
+---
+name: ts-react-frontend
+description: TypeScript React frontend workflow for components, state, routing, typecheck, tests, builds, and browser smoke evidence.
+---
+
+# TypeScript React Frontend
+
+## When To Use
+
+Use this skill when changing React components, hooks, routes, client state, styling, build configuration, or frontend tests.
+For user-visible changes, typecheck, build, and unit tests are not enough; include real-browser smoke or screenshot evidence before claiming acceptance.
+
+## Docs To Load
+
+- `Harness/workflows/ts-react-frontend.md`
+- `Harness/PROGRESS.md` and current task `tasks/<id>/PLAN.md`
+- Existing frontend README, package scripts, design system, and test setup.
+
+## Required Inputs
+
+- Screen, component, or flow being changed.
+- Expected behavior and acceptance criteria.
+- Existing package manager and verification commands.
+- Selector contract: stable accessible labels/roles and `data-testid` hooks for critical inputs, buttons, filters, rows, empty/error/loading states, and other targetable UI states.
+
+## Allowed Writes
+
+- Frontend source, styles, tests, and focused docs in task scope.
+- Generated evidence such as screenshots or test reports in existing artifact paths.
+- Current task `tasks/<id>/PLAN.md` when tracking a plan item.
+
+## Output Format
+
+Return changed files, UI behavior summary, commands run, browser or test evidence, and unresolved risks.
+Include selectors added or verified for CDP/Playwright/manual checks.
+
+## PROGRESS.md & Task PLAN.md Updates
+
+Update `Harness/tasks/<task-id>/PLAN.md` only when executing a tracked task or recording required validation evidence.
+
+## dispatch.md Usage
+
+Use `Harness/dispatch.md` when independent frontend tasks can run in parallel, such as components, tests, and browser checks.

package/templates/optional/skills/ui-ux-review/.claude/skills/ui-ux-review/SKILL.md

@@ -1,40 +1,40 @@
----
-name: ui-ux-review
-description: Screenshot-driven UI and UX review for responsive behavior, accessibility, visual polish, and task clarity.
----
-
-# UI/UX Review
-
-## When To Use
-
-Use this skill when reviewing or changing user-facing screens, layouts, visual hierarchy, interaction states, accessibility, or responsive behavior.
-
-## Docs To Load
-
-- `Harness/workflows/ui-ux-review.md`
-- `Harness/PROGRESS.md` and current task `tasks/<id>/PLAN.md`
-- Existing design system, component, or style documentation.
-
-## Required Inputs
-
-- Screens, routes, or components under review.
-- Target users and primary tasks.
-- Breakpoints, themes, and accessibility expectations.
-
-## Allowed Writes
-
-- UI code and style files already in scope for the task.
-- Screenshot or audit artifacts in existing evidence folders.
-- Current task `tasks/<id>/PLAN.md` when the review is part of a tracked plan.
-
-## Output Format
-
-Return prioritized findings with file or screen references, evidence, recommended fixes, commands run, and residual risks.
-
-## PROGRESS.md & Task PLAN.md Updates
-
-Update current task `tasks/<id>/PLAN.md` only for tracked review tasks or when recording required evidence.
-
-## dispatch.md Usage
-
-Use `Harness/dispatch.md` only for independent review streams such as separate routes or breakpoints.
+---
+name: ui-ux-review
+description: Screenshot-driven UI and UX review for responsive behavior, accessibility, visual polish, and task clarity.
+---
+
+# UI/UX Review
+
+## When To Use
+
+Use this skill when reviewing or changing user-facing screens, layouts, visual hierarchy, interaction states, accessibility, or responsive behavior.
+
+## Docs To Load
+
+- `Harness/workflows/ui-ux-review.md`
+- `Harness/PROGRESS.md` and current task `tasks/<id>/PLAN.md`
+- Existing design system, component, or style documentation.
+
+## Required Inputs
+
+- Screens, routes, or components under review.
+- Target users and primary tasks.
+- Breakpoints, themes, and accessibility expectations.
+
+## Allowed Writes
+
+- UI code and style files already in scope for the task.
+- Screenshot or audit artifacts in existing evidence folders.
+- Current task `tasks/<id>/PLAN.md` when the review is part of a tracked plan.
+
+## Output Format
+
+Return prioritized findings with file or screen references, evidence, recommended fixes, commands run, and residual risks.
+
+## PROGRESS.md & Task PLAN.md Updates
+
+Update current task `tasks/<id>/PLAN.md` only for tracked review tasks or when recording required evidence.
+
+## dispatch.md Usage
+
+Use `Harness/dispatch.md` only for independent review streams such as separate routes or breakpoints.