Skip to content

Conversation

@clutchski
Copy link

No description provided.

clutchski and others added 12 commits January 8, 2026 17:22
…detection

- Clarify three things to confirm with user: app, injection point, libraries
- Add three-part library discovery process: known integrations, docs search, SDK search
- Update reference files with comprehensive library detection lists
- Add documentation links for all languages (repos, hosted docs, package docs)
- Mark library lists as non-exhaustive with links to current docs

Co-Authored-By: Claude Opus 4.5 <[email protected]>
Grants Read, Glob, Grep, WebFetch permissions so the skill can
access its own reference files without prompting users.

Co-Authored-By: Claude Opus 4.5 <[email protected]>
Each reference file now has:
- Link to check latest version on package registry
- Instructions to install with pinned version number

Co-Authored-By: Claude Opus 4.5 <[email protected]>
Each reference file now shows how to programmatically find the latest version:
- Python: PyPI JSON API
- TypeScript: npm registry API
- Go: GitHub releases API
- Java: Maven Central search API
- Ruby: RubyGems API

Co-Authored-By: Claude Opus 4.5 <[email protected]>
- Add Step 2(d) to ask user for Braintrust project name
- Convert URLs to proper markdown links for better UX
- Add instructions to construct actual project URL with org/project name

Co-Authored-By: Claude Opus 4.5 <[email protected]>
- Add REQUIRED instruction to use WebFetch for version lookup
- Emphasize "Do NOT guess or use a version from memory"
- Add example PyPI API call for clarity

Co-Authored-By: Claude Opus 4.5 <[email protected]>
Test import before writing instrumentation code to catch
wrong function names (e.g., wrapOpenAI vs wrap_openai).

Co-Authored-By: Claude Opus 4.5 <[email protected]>
- Build app and fix any errors
- Run app to trigger LLM call
- Query Braintrust API to verify logs were received
- Debug and repeat until logs appear

Co-Authored-By: Claude Opus 4.5 <[email protected]>
Split into three focused skills:
- setup-braintrust-sdk: Main skill with installation + orchestration
- discover-sdk-integrations: Find LLM providers/frameworks
- install-sdk-integration: Add wrapper code and initialization

Added try-do-verify loop in Step 4 for verification.

Each skill has language-specific references (Python, TypeScript,
Go, Java, Ruby).

Co-Authored-By: Claude Opus 4.5 <[email protected]>
- Add step to ask user for run command or detect it
- Allow skill to either run itself or ask user to run
- Add explicit debug loop with repeat instruction
- Emphasize setup not complete until traces appear

Co-Authored-By: Claude Opus 4.5 <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants