What does this skill do, and when should you use it?
This skill is for developers upgrading @anthropic-ai/sdk or @anthropic-ai/claude-agent-sdk. It starts by checking installed versions and release notes, then consolidates version changes and inspects the public type surface. After installation, it guides compilation and Claude-related unit tests, followed by documentation and commit updates when needed. Its instructions are tailored to the Claude agent implementation in the vscode-copilot-chat project, making it a fit for controlled SDK migrations rather than a general dependency updater.
Reads current SDK versions from package.json and reviews the two Anthropic SDK release pages; organizes release changes into features, bug fixes, and breaking changes; updates both packages with npm install; snapshots .d.ts files into /tmp/anthropic-sdk-old before installation and uses diff afterward to identify new exports, parameters, signature changes, removals or renames, and deprecations; uses grep to cross-reference affected APIs with Claude agent code; runs npm run compile and Claude-related unit tests; and guides documentation updates, temporary-file cleanup, and a detailed upgrade commit message.
- A vscode-copilot-chat maintainer is moving @anthropic-ai/sdk to a newer release and needs a review of breaking API changes.
- A team using Claude Agent SDK needs to migrate between versions and consolidate all intervening release notes by category.
- An engineer is facing type errors after an SDK upgrade and needs to inspect exports, interfaces, signatures, and declaration-file changes.
- A reviewer needs to assess whether an SDK update affects session handling, message streaming, tools, hooks, or permission handlers.
- A project maintainer needs a post-upgrade compile and Claude test pass together with a commit message documenting versions and migration work.
How do you install this skill?
- Upgrading to “latest” may introduce unassessed breaking changes; define the target version and obtain confirmation before changes or commits.
- npm install and GitHub release-note review depend on external services; record resolved versions, lockfile changes, and recovery steps.
- Confirm ownership of /tmp/anthropic-sdk-old before rm -rf and prefer a recoverable cleanup approach.
- The skill does not explain whether sensitive source, environment variables, or SDK request data are sent to external services.
- Shell / CLI
- Network access
- Local filesystem
npm
The skill is located at extensions/copilot/.agents/skills/anthropic-sdk-upgrader/SKILL.md in the microsoft/vscode repository. The supplied material does not document a standalone installation command for other clients; it is one of 61 skills bundled in the repository. If the repository is available locally, read the SKILL.md at that path.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/microsoft/vscode.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/extensions/copilot/.agents/skills/anthropic-sdk-upgrader" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Upgrade the Anthropic SDK to the latest version
- Help me migrate to the latest claude-agent-sdk.
In a client that supports Agent Skills, make a request such as “Upgrade the Anthropic SDK to the latest version” or “Help me migrate to the latest claude-agent-sdk.” The documented workflow then operates around vscode-copilot-chat: inspect versions and release notes, run npm install, compare declaration files, check affected usage, and run npm run compile plus npm run test:unit -- --testPathPattern="agents/claude". The source does not document additional configuration for running it outside that project.
What are this skill's strengths and limitations?
- Covers version review, release-note consolidation, declaration-file diffs, usage cross-referencing, compilation, tests, and commit documentation.
- Separates findings into Critical, Important, and Nice to Have impact levels.
- Focuses on session, streaming, tool, hook, and permission-related Claude agent APIs.
- Provides concrete shell commands for creating a pre-upgrade snapshot and reviewing API-surface diffs.
- Its scope is limited to two Anthropic SDK packages rather than general dependency upgrades.
- The instructions target vscode-copilot-chat and name project-specific source paths, so other projects may require adaptation.
- No standalone installation command, automation script, or fixed SDK target version is provided.
- The workflow requires GitHub release-note access plus shell, filesystem, npm, compilation, and test commands; the source does not document testing on other environments.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Anthropic SDK Upgrade Assistant this page ✓ Microsoft · Official | 43 · Not recommended | ★ 194k | 11d ago | MIT |
| Azure AI Search for TypeScript ✓ Microsoft · Official | 50 · Use with care | ★ 2.7k | 3mo ago | MIT |
| GitHub Copilot SDK Upgrade Assistant ✓ Microsoft · Official | 28 · Not recommended | ★ 194k | 11d ago | MIT |
| Vibeyard: Add npm Dependency Skill | 55 · Use with care | ★ 1.4k | 1mo ago | MIT |
| Stable Interface Design | 50 · Use with care | ★ 107k | 12d ago | MIT |
How did FollowSkills review this skill?
The skill identifies the Anthropic SDK packages and describes the upgrade workflow, with no evidence of malware, credential theft, or covert exfiltration. However, it instructs the agent to run npm install, modify code, remove a fixed temporary directory, and create a commit without requiring user confirmation or documenting data flows, rollback, or recovery after failure, so points are deducted.
The workflow covers version review, type diffs, compilation, tests, and documentation, and the repository contains related CI and version-sync tests. However, the skill itself has no executable tests, assumes specific paths and files, uses an unspecified latest version, and lacks handling or diagnostic feedback for invalid input, network failures, and command failures; the static ceiling therefore applies.
Triggers, audience, and Anthropic SDK upgrade scenarios are clearly stated with several examples. Boundaries for non-fit cases, target-version constraints, permission scope, and Chinese-language guidance are absent, and GitHub release-note reachability from mainland China is not addressed, so the score is moderate.
The skill has a readable stepwise structure with package descriptions, an output format, troubleshooting guidance, command examples, and limitation hints. Repository evidence supplies an MIT license, Microsoft provenance, and maintenance signals, but the skill lacks its own version, changelog, explicit maintainer responsibility, or clear update path; prerequisites and rollback guidance are also incomplete.
The process could support SDK upgrades, API-difference analysis, compilation repair, and test verification, with a relatively clear report format. However, the target version is unspecified, the default is effectively latest, the process depends on external release notes, and it requires committing changes; outcomes still need human review, so static evidence supports only a limited score.
The supplied repository files provide dependency versions, relevant source paths, test commands, and limited CI evidence. There is no skill-specific test, execution record, or completed upgrade result, and no commands were executed in this review, so the key path is not independently reproducible.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →