Add mcp__mcp-docs and WebFetch to allowedTools so Claude can use
the MCP docs server and fetch schema content from GitHub URLs.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Cliff Hall <cliff@futurescale.com>
- Add .mcp.json with mcp-docs HTTP server configuration
- Update claude.yml to load MCP server config with --mcp-config
- Change --system-prompt to --append-system-prompt to preserve defaults
- Extend instructions to use mcp-docs for MCP-related work
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Address review feedback from @domdomegg to simplify the quickstart
links by removing the anchor to a specific step. Users should see
the full quickstart including verification, and simpler links are
more robust if docs reorganize.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Updated pull request template to inform contributors that adding servers to
the README is deprecated. Contributors should now publish servers to the MCP
Server Registry instead, following the quickstart guide.
Changes:
- Added "Publishing Your Server" section with deprecation notice
- Updated checklist to include MCP Registry publishing
- Clarified README item refers to server's own documentation
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
The 'Check if tests exist' step was actually running tests with
continue-on-error: true. If tests failed, it would set has-tests=false
and skip the actual test step, making CI appear green even with failing tests.
Simplified to use 'npm test --if-present' which:
- Runs tests if a test script exists (and fails if tests fail)
- Does nothing and exits 0 if no test script exists
- Removes the need for the complex check logic
Fixes the issue where PR #3014 had failing tests but CI was green.
Updates the Claude Code GitHub Action to use the stable v1 GA release instead of the beta version.
## Changes
- Updates action version from `@beta` to `@v1`
- Migrates `allowed_tools` to `claude_args: --allowedTools`
- Migrates `custom_instructions` to `claude_args: --system-prompt`
- Retains `additional_permissions` and `assignee_trigger` (both still supported in v1)
## Behavior
The action continues to work the same way:
- Triggers on `@claude` mentions in comments, reviews, and issues
- Triggers when assigned to an issue as "claude"
- Allows Claude to run Bash commands
- Allows Claude to read CI results on PRs
- Applies custom instructions for posting concise summaries
- in create-release job condition
- check for all combinations of publish results that should lead to a release creation
- npm = success / pypi = skipped
- npm = skipped / pypi = success
- npm = success / pypi = success
- in create-release job condition
- check for all combinations of publish results that should lead to a release creation
- npm = success / pypi = skipped
- npm = skipped / pypi = success
- npm = success / pypi = success
- in create-release job condition
- check for all combinations of publish results that should lead to a release creation
- npm = success / pypi = skipped
- npm = skipped / pypi = success
- npm = success / pypi = success
- in publish-pypi job,
- only proceed if there are python packages to publish
- in publish-npm job,
- only proceed if there are npm packages to publish
- Added test job that runs before build
- Tests are conditionally executed only for packages that have test scripts
- Filesystem tests will now run automatically in CI on every push and PR
- Build job now depends on successful test completion
