langfuse-core-workflow-a
Execute Langfuse primary workflow: Tracing LLM calls and spans. Use when implementing LLM tracing, building traced AI features, or adding observability to existing LLM applications. Trigger with phrases like "langfuse tracing", "trace LLM calls", "add langfuse to openai", "langfuse spans", "track llm requests".
64
77%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Passed
No known issues
Optimize this skill with Tessl
npx tessl skill review --optimize ./plugins/saas-packs/langfuse-pack/skills/langfuse-core-workflow-a/SKILL.mdQuality
Discovery
89%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is a solid description with excellent trigger terms and clear 'what/when' guidance. Its main weakness is that the capability description is somewhat thin—it mentions 'tracing LLM calls and spans' but doesn't enumerate more specific actions (e.g., decorating functions, creating generations, configuring scoring). The explicit trigger phrases and tool-specific naming make it highly distinctive and easy to select correctly.
Suggestions
Expand the capability list with more concrete actions, e.g., 'Traces LLM calls and spans, configures decorators for automatic tracing, logs generations and scores, integrates Langfuse with OpenAI/LangChain SDKs.'
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Names the domain (Langfuse, LLM tracing) and mentions some actions ('Tracing LLM calls and spans'), but doesn't list multiple concrete actions like creating traces, adding span metadata, configuring decorators, or viewing dashboards. | 2 / 3 |
Completeness | Clearly answers both 'what' (tracing LLM calls and spans via Langfuse) and 'when' (implementing LLM tracing, building traced AI features, adding observability) with explicit trigger phrases. | 3 / 3 |
Trigger Term Quality | Includes strong natural trigger terms: 'langfuse tracing', 'trace LLM calls', 'add langfuse to openai', 'langfuse spans', 'track llm requests'. These cover natural variations a user would actually say when needing this skill. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive due to the specific tool name 'Langfuse' and the narrow focus on LLM tracing/observability. Unlikely to conflict with other skills unless there's another Langfuse-specific skill. | 3 / 3 |
Total | 11 / 12 Passed |
Implementation
64%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a highly actionable skill with excellent, executable code examples covering multiple integration patterns for Langfuse tracing. Its main weaknesses are verbosity from including both v3 and v4 SDK patterns inline and the lack of validation checkpoints to verify traces are actually appearing in the dashboard. The error handling table is a nice touch but would be more effective if verification steps were woven into the workflow itself.
Suggestions
Add a validation checkpoint after Step 1 (e.g., 'Verify: Open Langfuse dashboard → Traces tab → confirm the trace appears with model, tokens, and latency') to ensure tracing is working before proceeding to complex patterns.
Move the v3 Legacy RAG pipeline (Step 3) to a separate reference file like `LEGACY-V3.md` and link to it, reducing the main skill's token footprint.
Consider moving the Anthropic and LangChain integration steps to separate files as well, keeping the core SKILL.md focused on the primary OpenAI wrapper and v4 manual tracing patterns.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is mostly efficient with executable code examples, but includes both v3 and v4 SDK versions inline which adds significant length. The v3 legacy section could be referenced externally. Some comments are unnecessary (e.g., 'Wrap the OpenAI client -- all calls are now traced automatically'). | 2 / 3 |
Actionability | Every step provides fully executable, copy-paste ready TypeScript/Python code with concrete examples covering OpenAI wrapper, manual tracing, streaming, Anthropic, and LangChain. The code includes specific models, metadata patterns, and token usage tracking. | 3 / 3 |
Workflow Clarity | Steps are clearly sequenced and labeled, but there are no validation checkpoints — no guidance on verifying traces appear in the Langfuse dashboard, no error recovery loops, and no verification that the SDK is properly initialized before tracing. The error handling table helps but is reactive rather than integrated into the workflow. | 2 / 3 |
Progressive Disclosure | The skill has good section structure and links to external resources, but includes too much inline content (especially the v3 legacy RAG pipeline and the Anthropic example) that could be split into separate reference files. With no bundle files, all ~180 lines of code live in one document. | 2 / 3 |
Total | 9 / 12 Passed |
Validation
81%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 9 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
allowed_tools_field | 'allowed-tools' contains unusual tool name(s) | Warning |
frontmatter_unknown_keys | Unknown frontmatter key(s) found; consider removing or moving to metadata | Warning |
Total | 9 / 11 Passed | |
- Commit
a04d1a2
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.