apps-mcp: Add integrated skills system for domain-specific guidance #4199
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Refactors init-template into dedicated subcommands with comprehensive L2 guidance for each resource type: - Add job subcommand using default-python template - Add pipeline subcommand using lakeflow-pipelines template - Add empty subcommand using default-minimal template - Create L2 templates: target_jobs, target_pipelines, target_mixed - Add AGENTS.tmpl for scaffolded projects - Parameterize default catalog in discover flow - Update detector to show "mixed" guidance for non-app-only projects - Move implementation to cmd/init_template/ subpackage - Make databricks_discover description more forceful (MANDATORY) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
…y configMap conditionals - Add acceptance tests for job, pipeline, empty, and app init-template subcommands - Remove unnecessary 'if catalog != ""' checks (templates handle empty values) - Fix logdiag double-initialization in bundle detector - Simplify FormatProjectScaffoldSuccess output formatting
- Remove undefined listAllSkills parameter from generateDiscoverGuidance call - Fix test expectation: jobs-only projects now include "mixed" target type 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
This makes the terminology more accurate for empty and single-resource bundles, and makes resource detection more maintainable. Changes: - Rename target_mixed.tmpl to target_bundle.tmpl - Use AllResources() iterator instead of manually checking each type - Simplify isAppOnly logic to: hasApps && len(TargetTypes) == 1 - Add default catalog logic to empty.go for consistency - Add test coverage for empty bundles 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove IsAppOnly field assignment (field was removed in base branch) - Remove verbose TestDetectorRegistry_AppsWithOtherResources test - Coverage is already provided by existing tests 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
lennartkats-db
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
Collaborator
|
Commit: 54c0a0e
24 interesting tests: 21 KNOWN, 2 FAIL, 1 SKIP
Top 33 slowest tests (at least 2 minutes):
|
The previous pattern `*/*` only matched one directory level deep, causing nested skill files like `pipelines/auto-cdc/SKILL.md` to be excluded from the embedded filesystem on Windows builds. Changed to `all:*` which: - Recursively embeds all subdirectories - Includes hidden files like .gitkeep - Automatically supports new resource type directories without code changes 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
lennartkats-db
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
The wildcard pattern `all:*` doesn't work reliably on Windows builds - the embedded filesystem ends up empty, causing all skills to be missing from the registry. Changed to explicit directory listing: `all:apps all:bundle all:jobs all:pipelines` This matches the pattern used in other parts of the codebase (e.g., libs/template/builtin.go uses `all:templates`). When adding new resource type directories, this list will need to be updated. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Adds TestAllSkillDirectoriesAreEmbedded which compares filesystem directories against the embedded FS and registry to ensure the //go:embed directive is complete. The test provides actionable error messages showing exactly which directories are missing and the correct embed directive to use. This prevents issues where new skill directories are added but not included in the explicit embed list (required for Windows compatibility). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
lennartkats-db
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
lennartkats-db
changed the title
Changed from space-separated pattern: //go:embed all:apps all:bundle all:jobs all:pipelines To separate directives (one per line): //go:embed all:apps //go:embed all:bundle //go:embed all:jobs //go:embed all:pipelines This matches the pattern used in libs/template/builtin.go and may be more robust on Windows. Also added panic messages with diagnostics to help identify the exact failure point if embed doesn't work. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
# Conflicts: # acceptance/apps/init-template/app/output.txt # acceptance/apps/init-template/app/script # acceptance/apps/init-template/empty/output.txt # acceptance/apps/init-template/empty/script # acceptance/apps/init-template/job/output.txt # acceptance/apps/init-template/job/script # acceptance/apps/init-template/pipeline/output.txt # acceptance/apps/init-template/pipeline/script # experimental/apps-mcp/cmd/init_template/app.go # experimental/apps-mcp/cmd/init_template/common.go # experimental/apps-mcp/lib/detector/bundle_detector.go # experimental/apps-mcp/lib/providers/clitools/discover.go
lennartkats-db
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
Updated regex to handle both Unix (\n) and Windows (\r\n) line endings. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
lennartkats-db
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
Tests would catch these issues anyway. Kept the actual error panics for fs.ReadDir and loadSkill failures. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
lennartkats-db
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Changes
This adds a skills system to apps-mcp that provides domain-specific implementation guides to AI agents. It follows up on #4183. Skills are organized by resource type and delivered through the discover tool. They work for any agent that supports MCP.
Detailed overview:
lib/skills/{apps,jobs,pipelines,...}and include a standardSKILL.mdfile with a skill front matter.databricks_discovertool.read_skill_filetool (which has a precedent in many other MCPs an in our first-party agent).lakeflow-mcpbranch; before merging it needs to be based onmaininsteadWhy
read_skill_filetool helps these agents use skills.Tests