本文へ移動
cccskills
無料GitHub で公開

pmstudio-changelog

Create or update a structured change log for a product or platform. Use when someone asks to "create a change log", "log this change", "what changed", "update the change log", "release notes", or "changelog". Reads project memory (CLAUDE.local.md Recent Changes) and meeting notes to build chronological change records. Supports init (backfill from memory), update (append new entries), and release (summarize for a version).

インストール方法を見る

含まれるファイル(1)

  • SKILL.md4.5 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。

Change Log — Running Product Change Record

Purpose

Maintains a chronological, append-only record of all meaningful changes to a product. Unlike CLAUDE.local.md (which is AI context), the change log is a stakeholder-facing document suitable for sharing with leadership, auditors, and team members.

Process

Step 1: Determine Mode

  • /change-log init — Create new change log, backfill from CLAUDE.local.md
  • /change-log update — Find changes since last entry, propose new entries
  • /change-log release {version} — Create release summary from accumulated entries
  • /change-log (no args) — Same as update

Step 2: Read Context

Always read:

  • CLAUDE.local.md — Recent Changes section (primary source of truth for what changed)
  • Existing change log file (if update or release mode)

For init mode, also read:

  • Meeting-Notes/ directory listing — to identify decision-bearing meetings
  • PRD/ — version history sections
  • .sync-watch.json — project name

For update mode:

  • Find the date of the last change log entry
  • Read any Meeting-Notes/ files dated after that
  • Read CLAUDE.local.md Recent Changes entries dated after that

Step 3: Extract Changes

For each source, extract structured change data:

- Date: YYYY-MM-DD
- Summary: one-line description
- Type: Feature | Enhancement | Bug Fix | Configuration | Process | Security | Documentation
- Impact: High | Medium | Low
- Components: which modules/layers/documents were affected
- Author: who made or requested the change
- Approved By: who approved (if known from meeting notes)
- Details: bullet list of specific changes
- Rationale: why the change was made
- Downstream: what else was updated as a result

Impact classification:

  • High — Changes product behavior, affects users, or modifies architecture
  • Medium — Updates documentation, adds stakeholders, changes processes
  • Low — Minor fixes, formatting, internal-only changes

Step 4: Propose Entries (update/init)

Present proposed entries to user before writing:

## Proposed Change Log Entries

### [2026-03-17] PRD v0.9 — User Journey Sections
**Type:** Enhancement | **Impact:** High
**Components:** ProductB PRD, ProductB PRD Presentation
- Added Current User Journey (6-step timeline)
- Added Future User Journey (3-layer cascading)
- Renumbered all sections (19 total)

Accept? [Y/n] or modify?

Step 5: Write

File location: Data/Change-Log-{ProjectName}.md

File structure:

# Change Log — {Product Name}

> Running record of all changes. Newest entries first.
> Generated and maintained via `/change-log` skill.

Last updated: {date}
Total entries: {count}

---

## [YYYY-MM-DD] v{X.Y} — {Summary}

**Type:** {type} | **Impact:** {impact}
**Components:** {affected components}
**Author:** {name} | **Approved By:** {name or "—"}

### Changes
- {specific change 1}
- {specific change 2}

### Rationale
- {why this change was made}

### Downstream Effects
- {what else was updated}

---

## [YYYY-MM-DD] ...

For release mode:

Collect all entries since the last release entry (or all entries if first release). Produce a summary:

## Release: v{version} — {date}

### Summary
{2-3 sentence overview of what this release includes}

### Changes Included
| Date | Type | Summary | Impact |
|------|------|---------|--------|
| ... | ... | ... | ... |

### Highlights
- {Most important change}
- {Second most important}

### Known Issues
- {Any caveats}

Critical Rules

  1. Append-only. Never modify existing entries. Only add new ones at the top.
  2. Propose before writing. Always show the user what entries will be added.
  3. Don't duplicate. Check existing entries before proposing. If a change is already logged, skip it.
  4. Real content only. Don't log trivial changes (typo fixes, formatting) unless the user specifically asks.
  5. Dates from sources. Use the date from CLAUDE.local.md or meeting notes, not today's date, for the change date. Today's date is for entries about changes made today.

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

Train and optimize AI agents using Microsoft's Agent Lightning framework with reinforcement learning. Use when setting up agent training, instrumenting agents with tracing, configuring LightningStore, implementing reward functions, or optimizing prompts with RL/APO algorithms.

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。

coco-research/coco5472026年10月11日 更新

Post-run self-evaluation system that scores agent output on correctness, clarity, actionability, and conciseness. Use after /team runs, skill executions, or when explicitly asked to evaluate output quality.

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。

coco-research/coco5472026年10月11日 更新

Create AI marketing videos for ads, promos, product launches, and brand content. Models: Veo, Seedance, Wan, FLUX for visuals, Kokoro for voiceover. Types: product demos, testimonials, explainers, social ads, brand videos. Use for: Facebook ads, YouTube ads, product launches, brand awareness. Triggers: marketing video, ad video, promo video, commercial, brand video, product video, explainer video, ad creative, video ad, facebook ad video, youtube ad, instagram ad, tiktok ad, promotional video, launch video

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。

coco-research/coco5472026年10月11日 更新

Use when building AI features into a product: LLM integration, RAG pipelines, guardrails, streaming, AI UX, prompt engineering, or AI cost control. Treats prompts as code and validates every model output.

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。

coco-research/coco5472026年10月11日 更新

Your AI research and engineering brain trust. 59 named personas across 8 cells covering frontier labs, applied product, model architecture, reasoning/RL/agents, alignment and interpretability, theory and science of DL, multimodal and…

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。

coco-research/coco5472026年10月11日 更新

Use when designing a new REST or GraphQL API, reviewing an API spec before implementation, setting team API standards, or migrating REST to GraphQL. Covers resources, HTTP semantics, pagination, error handling, and pitfalls.

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。

coco-research/coco5472026年10月11日 更新

coco-research のスキルをすべて見る

このスキルの問題を報告する