Claude Skill Changes Tracker
Pricing
Pay per usage
Claude Skill Changes Tracker
Daily diff watcher for Claude Code SKILL.md files across a curated list of GitHub repos. Detects new, updated, and removed Skills by comparing each run's manifest against a cached baseline in the Actor's key-value store.
Pricing
Pay per usage
Rating
0.0
(0)
Developer
Yanlong Mu
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
2 days ago
Last modified
Categories
Share
Daily diff watcher for Claude Code Skills across the official Anthropic repo and the top community repos. Find out — automatically, every day — which Skills were added, updated, or removed since yesterday.
What does Claude Skill Changes Tracker do?
The Claude Code Skills ecosystem moved from "interesting idea" to "production toolchain" in months. The official Anthropic Skills repo keeps growing, and dozens of community Skills repos publish their own SKILL.md files weekly. There is no central changelog.
This Actor solves that. It scans a configurable watchlist of GitHub repositories (default: the Anthropic-canonical + 4 top community Skills repos), finds every SKILL.md file in each repo's default branch, and diffs the manifest against the previous run. The result is a structured changelog: which Skills are brand new, which were updated (content hash or size changed), and which were removed.
It is the time-series companion to the point-in-time claude-code-skill-finder Actor (#13 in Ian Mu's 100-actor portfolio). When scheduled on the Apify platform, it gives you a daily changelog with zero infrastructure of your own — leveraging Apify's scheduling, key-value-store-backed snapshots, and API access.
Why use Claude Skill Changes Tracker?
- Skill maintainers: get a notification when a competing Skill releases a new version, or when the Anthropic official repo bumps
design-htmlorfrontend-design - Developer-tools investors: track the velocity of the Claude Code Skills ecosystem — is community contribution accelerating?
- Tutorial creators: spot brand-new Skills the day they land so your content is first
- Product managers at companies building on Claude Code: know the day a new official Skill changes your integration assumptions
- Security teams: audit removed Skills (deprecated for a reason?) and updated Skills (what changed?)
How to use Claude Skill Changes Tracker
- Open the Input tab. The default watchlist tracks 5 high-signal repos.
- (Optional) Add a
githubTokenApify secret to raise GitHub's anonymous rate limit (60/hour) to 5000/hour. - (Optional) Add or remove repos from
watchlist. - Click Start. The first run has no baseline — every Skill is reported as
newand the manifest is cached. - Schedule the Actor daily (e.g. 06:00 UTC). From the second run onward, the dataset returns only deltas: new, updated, removed.
- Pipe the dataset into Slack, email, or a database via Apify's webhooks / integrations.
Input
| Field | Type | Description |
|---|---|---|
watchlist | string[] | Array of owner/repo GitHub paths to monitor. Defaults to 5 curated Skills repos. |
cacheKeyPrefix | string | Prefix for per-repo manifests in the key-value store. Default skill-cache-. |
githubToken | string (secret) | Optional GitHub PAT to raise the API rate limit. |
branchHint | string | Force a non-default branch globally. Empty = auto-detect per repo. |
Example input:
{"watchlist": ["anthropics/skills","obra/superpowers","wshobson/agents"],"cacheKeyPrefix": "skill-cache-","githubToken": "ghp_xxx"}
Output
Each dataset item:
{"changeType": "updated","skillName": "design-html","skillPath": "skills/design-html/SKILL.md","repo": "anthropics/skills","branch": "main","sizeBytes": 4200,"previousSizeBytes": 3850,"sizeDeltaBytes": 350,"sha": "3a7c2d1...","previousSha": "1f9e8b4...","url": "https://github.com/anthropics/skills/blob/main/skills/design-html/SKILL.md","detectedAt": "2026-05-21T06:00:00.000Z"}
You can download the dataset in various formats such as JSON, HTML, CSV, or Excel. A Markdown summary report (skill-changes-report.md) is also written to the key-value store after every run.
Data table
| Field | Type | Meaning |
|---|---|---|
changeType | string | new, updated, or removed |
skillName | string | Directory name housing the SKILL.md (e.g. design-html) |
skillPath | string | Full path inside the repo (e.g. skills/design-html/SKILL.md) |
repo | string | owner/repo source |
branch | string | Branch the diff was computed against |
sizeBytes | number | Current SKILL.md byte size (null for removed) |
previousSizeBytes | number | Prior-run byte size (null for new) |
sizeDeltaBytes | number | Signed delta |
sha | string | Git blob SHA — proxy for content hash |
url | link | Direct GitHub link to the SKILL.md |
detectedAt | date | Timestamp of this run |
Pricing / Cost estimation
- Free trial: 5 daily runs, 5-repo watchlist (covers the default config)
- Daily delta subscription: $1.50/month for daily diffs of up to 10 repos
- Wide watchlist: $0.001 per repo scanned per run, so a 50-repo watchlist costs around $0.05/run
Each run uses around 0.001 Compute Units. The GitHub API is free up to 60 requests/hour anonymous and 5000/hour with a token.
Tips and advanced options
- First-run cost: every Skill is reported
newon run #1 because there is no baseline. Expect a one-time spike. - Pair with
vscode-ai-extension-popularity-tracker(sibling Actor in the 100-actor portfolio) to build an end-to-end "AI coding tool intelligence" pipeline - Add your own repo: if you maintain a Skills repo, include it in
watchlistfor a free daily audit of your own changes - Use webhooks: configure an Apify webhook on
RUN_SUCCEEDEDto pushchangeType=newrows directly into Slack - Forks vs upstream: this Actor does not deduplicate Skills that exist in both a fork and its upstream. If you want a deduped global Skills view, use
claude-code-skill-finder(the point-in-time companion).
FAQ, disclaimers, and support
Q: How does it detect "updated" vs "new"?
Each run records a manifest of { skillPath → { sizeBytes, sha } }. Next run compares the new manifest to the cached one. Any SHA or size change = updated. New path = new. Missing path = removed.
Q: Why filename + size + SHA, not full content? Cost. Fetching every SKILL.md body across 50 repos every day wastes bandwidth. The Git blob SHA is a perfect content hash — same SHA means byte-identical content.
Q: Will it miss content changes that don't change size? No. The Git blob SHA changes on any byte-level change, even ones that preserve size.
Q: What if a repo has 100+ SKILL.md files in subdirectories?
The tree API returns up to 100k entries. If truncated=true is reported, the Actor logs a warning. Practical Skills repos stay well under this cap.
Q: Can I track private repos?
Yes — pass a githubToken with appropriate scope via Apify secrets.
Q: Legality? This Actor only reads public GitHub metadata (repo tree + file size + SHA), which is explicitly allowed by GitHub's ToS for both authenticated and anonymous use.
For issues, feature requests, or custom watchlists, open the Issues tab on this Actor.
Built by Ian Mu — github.com/ianymu, author of claude-verify-before-stop. Part of a 100-Actor AI / dev-tools intelligence portfolio.