本文へ移動
cccskills
無料GitHub で公開

codex-project-pulse

Use when asking project status, next priority, sprint health, blockers, deadlines, risk register, or daily brief generation.

インストール方法を見る

含まれるファイル(5)

  • SKILL.md8.9 KB
  • agents/openai.yaml274 B
  • references/daily-brief-template.md6.7 KB
  • references/priority-engine.md6.2 KB
  • references/state-schemas.md9.1 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

TL;DR

Project dashboard brain. Maintains persistent state in .codex/pulse/ — sprint, priorities, blockers, risks, milestones. When user asks "hôm nay thế nào?" or "$today", read all state files, run quality check, and generate a comprehensive daily brief. This is the skill that makes the agent know WHERE the project is and WHAT to do next.

Activation

  1. Activate on $pulse, $today, $daily, $status, $brief.
  2. Activate on "hôm nay thế nào?", "status update", "what's next?", "what should I work on?", "project status", "daily standup", "morning brief".
  3. Activate on $sprint-init, $done, $block, $unblock, $milestone, $priority, $risk.
  4. Activate on "bắt đầu sprint", "hoàn thành story", "bị block", "deadline", "rủi ro".
  5. Auto-load with codex-workflow-autopilot when routing daily check-ins.
  6. Auto-load with codex-scrum-subagents during sprint ceremonies.
  7. Auto-suggest on session start if .codex/pulse/sprint-state.json exists.

Announce at start: "Loading project pulse — checking sprint state, priorities, and blockers."

State Files

All project state lives in .codex/pulse/ at the project root:

.codex/pulse/
├── sprint-state.json       # Current sprint: goal, stories, status, dates
├── priority-queue.json     # Ordered list of what to do next
├── blockers.json           # Active blockers with owners and age
├── risk-register.json      # Risks with severity, probability, mitigation
├── milestones.json         # Deadlines and milestone tracking
└── daily-brief.md          # Latest generated morning brief

Command Reference

Daily Operations

CommandActionExample
$today / $pulseGenerate daily brief"Hôm nay thế nào?"
$priorityShow priority queue"Việc gì cần làm trước?"
$priority reorderReorder priorities"Đổi thứ tự ưu tiên"

Sprint Management

CommandActionExample
$sprint-initInitialize new sprint"$sprint-init Sprint 4 --goal 'Complete payment flow' --end 2026-05-10"
$sprint-closeClose current sprint, calculate velocity"$sprint-close"
$add-storyAdd story to sprint backlog"$add-story PAY-001 'Stripe checkout' --points 5 --owner @dev-a"
$done <id>Mark story complete"$done PAY-001"
$wip <id>Mark story in-progress"$wip PAY-002"

Blocker Management

CommandActionExample
$block <id> <reason>Add blocker"$block PAY-001 'Waiting for Stripe API key'"
$unblock <id>Remove blocker"$unblock PAY-001"
$blockersList all active blockers"$blockers"

Risk & Milestone Management

CommandActionExample
$risk <desc>Add risk"$risk 'Payment integration delay' --severity high"
$risk resolve <id>Close a risk"$risk resolve RISK-003"
$milestone <name> <date>Add milestone"$milestone 'Demo day' 2026-05-15"
$milestonesShow all milestones with proximity"$milestones"

Daily Brief Protocol

When $today or $pulse is triggered:

Step 1: Read State

1. Read .codex/pulse/sprint-state.json → sprint progress
2. Read .codex/pulse/priority-queue.json → what's next
3. Read .codex/pulse/blockers.json → what's stuck
4. Read .codex/pulse/risk-register.json → active risks
5. Read .codex/pulse/milestones.json → upcoming deadlines

Step 2: Gather Signals

6. Check git log --since=yesterday → recent commits
7. Run quality check if project has test/lint commands
8. Read latest .codex/sessions/ summary → yesterday's context
9. Count open TODOs/FIXMEs if available

Step 3: Generate Brief

10. Calculate sprint progress (stories done / total, time elapsed / total)
11. Identify overdue milestones
12. Identify stale blockers (> 2 days old)
13. Generate priority recommendations
14. Write .codex/pulse/daily-brief.md
15. Present to user

Step 4: Recommend Action

16. If blockers exist → recommend unblocking action first
17. If deadline < 3 days → flag urgency
18. If quality degrading → recommend quality fix before features
19. Otherwise → recommend top priority from queue

Daily Brief Output Format

# 📊 Daily Brief — [Date]

## Sprint [N]: [Goal]
**Progress:** [X/Y stories] ([Z]%) | **Time:** Day [D/T] ([P]%)
**Velocity risk:** [On track / Behind / Ahead]

[Progress bar visualization]

## 🔴 Blockers ([count])
| Story | Blocker | Age | Owner | Action Needed |
|---|---|---|---|---|
| [id] | [reason] | [days] | [who] | [next step] |

## ⚠️ Risks ([count active])
| Risk | Severity | Probability | Mitigation |
|---|---|---|---|
| [desc] | 🔴/🟡/🟢 | H/M/L | [action] |

## ⏰ Upcoming Deadlines
| Milestone | Date | Days Left | Status |
|---|---|---|---|
| [name] | [date] | [N days] | 🟢/🟡/🔴 |

## 📈 Quality Pulse
| Signal | Status | Trend |
|---|---|---|
| Tests | [pass rate] | [↑/↓/→] |
| Lint | [error count] | [↑/↓/→] |
| Tech debt | [signal count] | [↑/↓/→] |
| Security | [finding count] | [↑/↓/→] |

## 📋 Today's Priorities
1. 🔴 [Urgent: unblock/fix] — [reason]
2. 🟡 [High: continue/start] — [story or task]
3. 🟢 [Normal: next in queue] — [story or task]

## 💡 Yesterday's Progress
[Summary from latest session or git log]

Sprint State Management

Sprint Lifecycle

stateDiagram-v2
    [*] --> Planning: $sprint-init
    Planning --> Active: Stories added
    Active --> Active: $done, $wip, $block
    Active --> Closing: All stories done or sprint end date
    Closing --> Review: $sprint-close
    Review --> [*]: Velocity recorded

    Active --> Active: Daily $pulse

Velocity Tracking

After each sprint close, record:

{
  "sprint": "Sprint 3",
  "planned_points": 21,
  "delivered_points": 18,
  "velocity": 18,
  "stories_planned": 7,
  "stories_completed": 6,
  "carryover": ["PAY-003"]
}

Use velocity history to forecast future sprint capacity:

  • recommended_commitment = average(last 3 sprints velocity) × 0.85

Priority Calculation Rules

When ordering the priority queue:

Priority Score = base_priority + urgency_bonus + blocker_penalty + deadline_bonus

base_priority:
  - critical: 100
  - high: 70
  - medium: 40
  - low: 10

urgency_bonus:
  - blocked_dependency: +30 (something else waiting on this)
  - stale_wip > 3 days: +20

blocker_penalty:
  - currently_blocked: -50 (deprioritize until unblocked)

deadline_bonus:
  - milestone < 3 days: +40
  - milestone < 7 days: +20
  - milestone < 14 days: +10

State Initialization

When no .codex/pulse/ exists, on $sprint-init:

  1. Create .codex/pulse/ directory
  2. Ask user for:
    • Sprint name and number
    • Sprint goal (one sentence)
    • Start date and end date
    • Stories to include (or import from backlog)
  3. Generate initial state files
  4. Generate first daily brief

When .codex/pulse/ already exists but sprint is closed:

  1. Archive previous sprint to .codex/pulse/archive/sprint-N/
  2. Initialize new sprint state
  3. Carry over incomplete stories (if user confirms)

Integration with Other Skills

SkillIntegration
codex-project-memoryRead session summaries for "yesterday's progress"; read decisions for context
codex-execution-quality-gateRun quality scripts for "quality pulse" section of daily brief
codex-scrum-subagentsUse ceremony workflows for sprint events; share sprint state
codex-workflow-autopilotRoute $today to pulse; use priority queue for "what next?" decisions
codex-context-engineLoad genome.md for project structure awareness alongside sprint state
codex-verification-disciplineVerify story completion claims with evidence before marking $done
codex-document-writerUse report templates for sprint reports and progress updates

Hard Rules

  • Never mark a story done without verification evidence (tests, build, demo).
  • Never remove a blocker without confirming the blocking condition is resolved.
  • Never ignore a milestone < 3 days away — always flag in daily brief.
  • Always read state files before generating any brief — no guessing from memory.
  • Always persist state changes to files — memory alone decays between sessions.
  • Stale blockers (> 2 days) get escalation flag automatically.
  • Sprint velocity is calculated from actual delivered points, not planned.
  • When quality degrades (test failures, new critical debt), flag before feature work.

Reference Files

  • references/state-schemas.md: JSON schemas for all .codex/pulse/ state files.
  • references/daily-brief-template.md: daily brief format, quality rules, and examples.
  • references/priority-engine.md: priority calculation, reordering rules, and queue management.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work — guides completion of development work by presenting structured options for merge, PR, or cleanup

日本語の概要は準備中です。原文の説明を表示しています。

Bang-isme/CodexAI---Skills102026年10月5日 更新

Use when a project has 50+ files, missing context, or explicit $codex-genome request; generates multi-layer project context to reduce hallucination.

日本語の概要は準備中です。原文の説明を表示しています。

Bang-isme/CodexAI---Skills102026年10月5日 更新

Use for DESIGN.md visual identity contracts; author, scaffold, lint, diff, and export design contracts with the upstream spec and CLI.

日本語の概要は準備中です。原文の説明を表示しています。

Bang-isme/CodexAI---Skills102026年10月5日 更新

Use when the task involves reading, creating, or editing `.docx` documents, especially when formatting or layout fidelity matters; prefer `python-docx` plus the bundled `scripts/render_docx.py` for visual checks.

日本語の概要は準備中です。原文の説明を表示しています。

Bang-isme/CodexAI---Skills102026年10月5日 更新

Use when code changes may require documentation updates; maps git diff to candidate docs without auto-editing documentation.

日本語の概要は準備中です。原文の説明を表示しています。

Bang-isme/CodexAI---Skills102026年10月5日 更新

Use for documents, reports, memos, guides, academic papers, or formal text that needs clear structure, reliable tone, formatting, and diagrams.

日本語の概要は準備中です。原文の説明を表示しています。

Bang-isme/CodexAI---Skills102026年10月5日 更新

Bang-isme のスキルをすべて見る

このスキルの問題を報告する