progress-dashboard
Use when asked to "give this session a dashboard", "show loop progress", "monitor subagents", or "steer this run in a browser": a durable local workspace for current work, findings, agents, measured usage and feedback.
- License 许可
- MIT
- 最近更新
- 2026-11-14
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- progress-dashboard
- metadata
- { "owner": "tze", "status": "active", "authors": [ "tze", "OpenAI Codex", "Anthropic Claude" ], "last_reviewed": "2026-11-14" }
- description
- Use when asked to "give this session a dashboard", "show loop progress", "monitor subagents", or "steer this run in a browser": a durable local workspace for current work, findings, agents, measured usage and feedback.
- compatibility
- Python 3.11+ and uv; stdlib-only server on Linux/macOS. Codex uses its local runtime relationship index; explicit Codex/Claude JSONL attachment remains available. No frontend build or model API.
02 资源文件 15
- 参考文档agents/openai.yaml289 B
- 脚本assets/web/app.js33.5 KB
- 资产文件assets/web/index.html5.7 KB
- 资产文件assets/web/style.css15.9 KB
- 参考文档references/artifacts.md1.1 KB
- 参考文档references/lifecycle.md4.0 KB
- 参考文档references/protocol.md6.1 KB
- 参考文档references/telemetry.md3.9 KB
- 脚本scripts/dashboard_server.py12.5 KB
- 脚本scripts/loop_telemetry.py19.2 KB
- 脚本scripts/progress_dashboard.py32.1 KB
- 脚本scripts/runtime_sources.py7.1 KB
- 脚本tests/test_dashboard_browser.py20.8 KB
- 脚本tests/test_loop_telemetry.py18.3 KB
- 脚本tests/test_progress_dashboard.py32.7 KB
03 SKILL.md 正文
Progress Dashboard
Publish a running workflow's progress to ~/.tmp/<slug>/ and a local page. The
helper owns files, layout, charts, usage collection and the server; supply only
short changed content. The owning tracker, claims and gates stay authoritative.
Start or resume
Resolve DASHBOARD to scripts/progress_dashboard.py
and pick one stable slug per parent session. Keep only that path and slug across
compaction; the workspace retains revision and feedback position. Load this skill once.
uv run "$DASHBOARD" --slug "$SLUG" start --title "Accepted outcome" --outcome "$TRACKER_ID"
start creates or resumes the workspace and server and prints the private URL;
report it, and keep it out of commits/PRs. Repeating it is safe. Remote access,
custom identity, recovery: lifecycle. Token usage is
measured automatically for Codex; other runtimes: usage.
Publish deltas at existing checkpoints
Publish when something meaningful changes: admission, delegation, a finding, a wait, completion. Reuse tracker reads and native agent results you already have. Never add dashboard-only agents, transcript summaries, timers or wakeups.
uv run "$DASHBOARD" --slug "$SLUG" checkpoint <<'JSON'
{"phase":"review",
"tasks":[{"id":"task-1","status":"complete","evidence":"PR 12 merged; 31 tests pass."},
{"id":"task-2","title":"Verify recovery","status":"in-progress"}],
"agents":[{"id":"reviewer-1","name":"Reviewer","role":"review","task":"task-2",
"summary":"Checking failure and recovery paths.","activity":["Started review."]}],
"findings":["Recovery survives a restart."]}
JSON
Rows merge by id: send only new rows and the changed fields of existing ones.
Never resend the roster. A new task needs title; statuses are
in-progress|todo|complete (tasks) and running|waiting|complete|failed|unknown
(agents). findings and agent activity append. summary is the one current
paragraph. Link an agent to its task with task; add branch, worktree and
pr_url when known. The page derives counts, charts, links and the activity
log; never write markup. For a chart or table, send rows to a fixed
panel template; for files, see
artifacts.
The reply is {revision, pending_count, feedback, ...}. Do not reread generated
files or copy state into chat. With nothing to publish, inbox reads feedback.
Feedback and authority
Each reply carries feedback not yet shown to this session. It waits for your next checkpoint and cannot run commands, change the tracker or approve a gate. After evaluating an item, record what you did:
uv run "$DASHBOARD" --slug "$SLUG" ack --id 3 --response "Accepted; verifying recovery next."
A truncated item must be read in full (inbox --id N) before acting. Formal
owner decisions use user-questionnaire.
Details: protocol.
Finish
Publish the final state, then uv run "$DASHBOARD" --slug "$SLUG" stop; content
is retained. handover.md changes only on an explicit owner request through the
manual handover command.
04 同仓库更多 Skills
spec-concretization
Use when specs and implementation must be fully reconciled in both directions — code doing things no spec describes, specs requiring things the code doesn't do yet, or the two disagreeing — and the result must land as vision-gated spec updates plus a complete, backlog-deduplicated work plan. Also the read-only audit form ("does code match the spec?"), and re-runs that must not duplicate prior findings. Triggers: "reconcile spec vs implementation", "does code match the spec", "what's implemented but undocumented", "what's specified but not built", "the code is ahead of the spec", "plan the work to finish the spec", "concretize the specs".
skill-creator
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.
frontend-design
Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, artifacts, posters, or applications (examples include websites, landing pages, dashboards, React components, HTML/CSS layouts, or when styling/beautifying any web UI). Generates creative, polished code and UI design that avoids generic AI aesthetics.
webapp-testing
Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.
web-artifacts-builder
Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn/ui). Use for complex artifacts requiring state management, routing, or shadcn/ui components - not for simple single-file HTML/JSX artifacts.
theme-factory
Toolkit for styling artifacts with a theme. These artifacts can be slides, docs, reportings, HTML landing pages, etc. There are 10 pre-set themes with colors/fonts that you can apply to any artifact that has been creating, or can generate a new theme on-the-fly.