| name | ai-changelog |
| description | Scan for new AI coding tool features and assess impact on your project. Covers Claude Code, Cursor, Copilot, Windsurf, Codex, Roo, and Goose. Use to track AI tooling changes and identify workflow improvements. |
| user-invocable | true |
| argument-hint | [scan|briefing|impact <feature>] [--tools claude,cursor] |
| allowed-tools | WebSearch, WebFetch, Read, Write, Grep, Glob |
| tags | ["meta"] |
AI Changelog
Scan AI coding tool changelogs and blogs, synthesize findings, and assess impact on your project's workflows.
Tool Filtering
If the user specifies --tools, only scan and report on those tools. Accepts comma-separated names.
Valid tool names: claude, cursor, copilot, windsurf, codex, roo, goose
Examples:
/ai-changelog scan --tools claude,cursor — only Claude Code and Cursor/ai-changelog briefing --tools claude — only Claude Code
If no --tools flag is provided, auto-detect from the project: look for .claude/, .cursor/, .github/copilot-instructions.md, etc. If nothing is detected, scan all tools.
Release Maturity
When scanning sources that distinguish release stages, classify each finding:
| Stage | Meaning | Tag |
|---|
| Stable | GA release, available to all users | (none) |
| Experimental | Pre-release or experimental flag, opt-in only | [EXP] |
| Beta/Preview | Public beta or preview program | [BETA] |
How this affects each mode:
- Scan: Tag pre-release findings with
[EXP] or [BETA] in the bullet list
- Briefing: Include a "Release Stage" line in each per-tool finding
- Impact: Adjust recommendation language — experimental features use "prepare for / monitor" rather than "adopt now"
GitHub Releases API entries with "prerelease": true are Experimental. Blog posts announcing "beta" or "preview" programs are Beta/Preview. Everything else is Stable.
Mode Selection
Determine mode from the argument:
1. Scan (/ai-changelog scan)
Quick scan for recent AI tooling news.
- Read
references/sources.md for the curated source list
- Filter to selected tools only (from
--tools flag or auto-detected)
- WebSearch each matching source category for posts from the last 30 days
- Output raw findings as a bullet list, grouped by tool
2. Briefing (/ai-changelog briefing)
Full research with structured synthesis.
- Run scan (step 1 above)
- WebFetch the most relevant results for detail
- Understand the project context:
- Read the project's CLAUDE.md, README, or equivalent for an overview
- Scan for AI workflow files (skills, rules, agents, hooks, .cursorrules, copilot instructions)
- Identify the AI tools the project targets
- For each significant finding, assess project impact using
references/impact-template.md
- Synthesize into a structured briefing:
- Per-tool sections — What changed, when, significance
- Impact assessment — Which project workflows are affected
- Recommendations — Concrete next steps (new workflow, update existing, no action)
- Save briefing to
briefings/YYYY-MM-DD.md
- Present summary to user
3. Impact (/ai-changelog impact <feature>)
Analyze a specific feature or change against the current project.
- Scan the project for AI workflow files and configuration
- WebSearch for details about the specified feature
- Assess which project workflows are affected
- Output using the impact template
Agent Mode
When invoked with --agent flag (for GHA automation):
- Skip interactive prompts
- Run full briefing mode
- Save to
briefings/YYYY-MM-DD.md
- Output JSON summary to stdout
References
| File | Purpose |
|---|
references/sources.md | Curated list of changelog URLs and blogs to scan |
references/impact-template.md | Template for project impact assessment |
