ml-pipeline
Skill2 starsChanged 6 months agoMITView on GitHub
ML pipeline orchestrator — single entry point for ML-related tasks. Coordinates ML Engineer (analysis, modeling, recommendations), SRE Engineer (production data extraction), and Product Manager (task formulation). Domain context from AGENTS.md. MVP flow: define data requirements → extract from prod → analyze → model → recommend → feature plan.
13 other projects use a version of this.
Adopted by
Most recently taken up by
- lamb92009/claude-skillstheir own version, first seen Nov 16, 2026
- santiago-migoni/claude-skillstheir own version, first seen Nov 8, 2026
- lsyamsiar/claude-skillstheir own version, first seen Nov 1, 2026
- codeApe-7/ai-agent-workflowGrouptheir own version, first seen Nov 1, 2026
- hu465/claude-skills-their own version, first seen Oct 25, 2026
- zacklecon/claude-skillstheir own version, first seen Oct 13, 2026
Used with
Projects with it also use ui-ux-pro-max.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
StrayDragonSkillui-ux-pro-max
UI/UX design intelligence. 50 styles, 21 palettes, 50 font pairings, 20 charts, 8 stacks (React, Next.js, Vue, Svelte, SwiftUI, React Native, Flutter, Tailwind). Actions: plan, build, create, design, implement, review, fix, improve, optimize, enhance, refactor, check UI/UX code. Projects: website, landing page, dashboard, admin panel, e-commerce, SaaS, portfolio, blog, mobile app, .html, .tsx, .vue, .svelte. Elements: button, modal, navbar, sidebar, card, table, form, chart. Styles: glassmorphism, claymorphism, minimalism, brutalism, neumorphism, bento grid, dark mode, responsive, skeuomorphism, flat design. Topics: color palette, accessibility, animation, layout, typography, font pairing, spacing, hover, shadow, gradient.
Also in 3 projects that use this
Other versions13 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 13
- Jeffallan/claude-skills★ 12ksame name
- nimadorostkar/Claude-Skills-collection★ 37same name
- codeApe-7/ai-agent-workflowGroup★ 18same name
- HyperFrequency/skills★ 15same name
- PIXARTSeu/Synapse★ 9same name
- zacklecon/claude-skills★ 3same name
- labdaps/labskills★ 2same name
- hu465/claude-skills-★ 0same name
- james-paul25/claude-skill★ 0same name
- lamb92009/claude-skills★ 0same name
3 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository40 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- ai-skillsSkill
- analyze-localSkill
- analyze-prodSkill
- analyzeSkill
- architectureSkill
- asset-validationSkill
- blog-postSkill
- bugfixSkill
- cloud-platformsSkill
- code-reviewSkill
- content-creationSkill
- context-engineeringSkill
- create-prSkill
- deployment-proceduresSkill
- deploy-productionSkill
- deploy-stagingSkill
- docsSkill
- feature-devSkill
- geo-writerSkill
- humanizerSkill
- infra-changeSkill
- marketing-operationsSkill
- marketingSkill
- planSkill
- pre-commitSkill
- productSkill
- project-initSkill
- prompt-engineeringSkill
- qaSkill
- releaseSkill
- run-testsSkill
- security-scanSkill
- seo-reviewSkill
- social-media-managerSkill
- test-localSkill
- test-strategySkill
- ui-ux-designSkill
- worktree-isolationSkill
The file
We split this skill into the rules it gives an agent
7 rules. 0 are common, 4 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 11 edited versions we read keep these.
4
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 27
If the user provides a vague request — proceed to Step 2 where product-manager role will help formulate it.
- Distinctive: line 46
Read and internalize the project's domain:
- Distinctive: line 56
Current prompts: If the product uses LLM prompts for data collection — identify prompt locations and purpose
- Very specific: line 125
Follow analyze-prod skill Steps 3a–3b to verify cloud platform context.
- Distinctive: line 144
Verify data completeness — row counts, date ranges, null percentages
- Distinctive: line 170
Document findings and assumptions
- Very specific: line 260
After user approves the recommendations, invoke plan skill with the following input:
“Common” is measured: kept by at least 6 in 10 of the 11 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
ML Pipeline
Single entry point for ML-related tasks. Orchestrates three roles across a data-driven analysis pipeline: Product Manager formulates the task, ML Engineer owns analysis and modeling, SRE Engineer extracts production data. Domain specifics are read from the project's AGENTS.md.
Architecture note: This workflow is extensible — future sub-flows (model training, prompt tuning, A/B testing) can be added as numbered branches in Step 2.
1. Receive ML Task
Gather the ML task from the user:
- Objective: What do you want to learn or improve? (e.g., "tune scoring weights", "improve data collection prompts", "analyze conversion funnel")
- Data source: Where is the data? (production database, logs, external API, local files)
- Expected output: What decisions will the analysis inform?
- Constraints: Timeline, data sensitivity, compute budget
If the user provides a vague request — proceed to Step 2 where product-manager role will help formulate it.
2. Determine Pipeline Type
Select the pipeline based on the ML task:
| Pipeline Type | When to Use | Status |
| Data Analysis | Extract prod data → analyze → model → recommend parameter/prompt changes | ✅ Active |
| Model Training | Train or retrain ML models on collected data | 🔮 Future |
| Prompt Tuning | Optimize LLM prompts based on output quality analysis | 🔮 Future |
| A/B Test Analysis | Analyze results of live experiments | 🔮 Future |
For MVP, proceed with Data Analysis pipeline (Steps 3–9). Future pipeline types will be added as parallel branches from this step.
3. Read Project Context
Apply product-manager role.
Read and internalize the project's domain:
AGENTS.md(root) — tech stack, domain description, business contextARCHITECTURE.md— system components, data flow, service boundariesFEATURES.md(if exists) — current feature inventory, scoring/evaluation parameters
Extract and summarize:
- Domain: What does the product do? What data does it collect?
- Evaluation parameters: What metrics/weights/scores does the product use?
- Data sources: Where is production data stored? (database, logs, analytics)
- Current prompts: If the product uses LLM prompts for data collection — identify prompt locations and purpose
Present the domain summary to the user for confirmation before proceeding.
4. Formulate ML Task
Continue with product-manager role.
Transform the user's request into a structured ML task specification:
## ML Task Specification
### Objective
[What we want to learn or improve — measurable outcome]
### Business Context
[Why this matters — which product decisions depend on this analysis]
### Success Criteria
[How we know the analysis is complete and actionable]
- [ ] Data requirements defined and extracted
- [ ] Statistical analysis performed with confidence intervals
- [ ] Recommendations backed by evidence (not intuition)
- [ ] Action plan produced (parameter changes, prompt modifications, or feature work)
### Scope
- **In scope**: [specific parameters, data ranges, user segments]
- **Out of scope**: [what we explicitly skip]
### Constraints
- Data sensitivity: [PII handling, anonymization needs]
- Time range: [how far back to analyze]
- Compute budget: [any limits on processing]Present to user for approval. Adjust based on feedback.
5. Define Data Requirements
Apply ml-engineer role.
Based on the ML task specification from Step 4, define exactly what data is needed:
<data_requirements>
- Data points: Which fields/columns/metrics are needed?
- Granularity: Per-record, aggregated, time-series?
- Time range: From when to when?
- Filters: Which segments, categories, statuses?
- Volume estimate: Expected row count / data size
- Format: CSV, JSON, Parquet, database query result?
- Joins: Which tables/sources need to be combined?
- Anonymization: Which fields contain PII and how to handle? </data_requirements>
Produce a concrete data extraction plan — either SQL queries, API calls, or log extraction commands that sre-engineer role can execute.
Present the data requirements and extraction plan to the user for review.
6. Extract Production Data
Apply sre-engineer role.
Execute the data extraction plan from Step 5. This step uses SRE's production access and operational expertise.
⚠️ SAFETY: All commands are READ-ONLY. No mutations to production data or systems.
6a. Establish Production Access
Follow analyze-prod skill Steps 3a–3b to verify cloud platform context. Consult cloud-platforms skill for platform-specific CLI commands.
6b. Execute Data Extraction
Based on the extraction plan:
| Data Source | Method |
| Managed DB (SQL) | Platform-specific connection (see cloud-platforms skill CLI Reference → Data Extraction) + psql/mysql query |
| Application logs | Platform-specific log query (see cloud-platforms skill Observability → Diagnostic Commands) |
| Monitoring metrics | Platform-specific metrics API (see cloud-platforms skill Observability) |
| Data warehouse | Platform-specific query tool (BigQuery / Synapse / Athena) |
| Object storage | Platform-specific CLI (gsutil / az storage / aws s3) |
| Kubernetes pod data | kubectl logs or kubectl exec (read-only) |
For each extraction:
- Show the exact command/query to the user before execution
- Execute after approval
- Save results to a local working directory (e.g.,
data/raw/) - Verify data completeness — row counts, date ranges, null percentages
6c. Data Handoff
Present extraction summary:
## Data Extraction Summary
- Source: [database/logs/metrics]
- Records extracted: [count]
- Time range: [from — to]
- Files: [local paths]
- Data quality notes: [nulls, anomalies, gaps]7. Analyze Data and Build Models
Apply ml-engineer role.
Perform the analysis following ML Engineer's reasoning protocol:
7a. Exploratory Data Analysis (EDA)
- Profile the extracted data: distributions, correlations, outliers, missing values
- Visualize key patterns (histograms, scatter plots, time series, heatmaps)
- Identify data quality issues that could affect analysis
- Document findings and assumptions
7b. Statistical Analysis
- Hypothesis testing for differences between segments/groups
- Correlation analysis between parameters and outcomes
- Time-series analysis for trends and seasonality (if applicable)
- Confidence intervals for all key metrics — never report point estimates without uncertainty
7c. Modeling (if applicable)
- Establish a baseline (current parameter weights, simple heuristic, or mean prediction)
- Train models to identify optimal parameters: regression, classification, or optimization
- Evaluate against baseline with appropriate metrics
- Feature importance analysis — which parameters matter most?
- Cross-validation for robust estimates
7d. Document Results
Present analysis as a structured report:
## Analysis Report
### Data Summary
- Records analyzed: [count]
- Time period: [range]
- Segments: [if applicable]
### Key Findings
1. [Finding with statistical evidence]
2. [Finding with statistical evidence]
3. [Finding with statistical evidence]
### Visualizations
[Charts, plots, tables illustrating findings]
### Model Performance (if applicable)
- Baseline: [metric = value]
- Proposed: [metric = value]
- Improvement: [delta with confidence interval]8. Produce Recommendations
Continue with ml-engineer role, consult product-manager role for business framing.
Translate analysis findings into actionable recommendations:
## ML Recommendations
### Summary
[One paragraph: what the analysis revealed and what should change]
### Recommended Changes
#### Parameter/Weight Changes
| Parameter | Current Value | Recommended Value | Expected Impact | Confidence |
|---|---|---|---|---|
| [param] | [current] | [recommended] | [metric improvement] | [high/medium/low] |
#### Prompt Modifications (if applicable)
| Prompt | Current Issue | Recommended Change | Expected Improvement |
|---|---|---|---|
| [prompt location] | [what's wrong] | [specific edit] | [quality metric] |
When recommendations include prompt or context pipeline changes (RAG, memory, agent harness, token budget), consult `context-engineering` skill for architecture patterns and production checklists.
#### Other Recommendations
- [Infrastructure, data pipeline, monitoring changes]
### Evidence
[Reference to analysis report, key metrics, statistical tests]
### Risks
| Risk | Mitigation |
|---|---|
| [what could go wrong] | [how to handle it] |
### Validation Plan
- [ ] A/B test for [specific changes] over [duration]
- [ ] Monitor [metrics] for regression
- [ ] Rollback criteria: [threshold]Present recommendations to the user. Wait for approval before proceeding.
9. Generate Feature Plan
After user approves the recommendations, invoke plan skill with the following input:
Feature specification (auto-generated from Step 8):
- Goal: Implement ML-recommended changes to [parameters/prompts/models]
- Requirements: Each recommended change from Step 8 becomes a functional requirement
- Acceptance criteria: Metrics improve by the predicted amounts (with tolerance)
- Constraints: Include validation plan (A/B test, monitoring) as non-functional requirements
plan skill will decompose the work into role-scoped work packages and produce an implementation plan.
10. Summary
Present the completed ML pipeline run:
- ML task: original objective
- Domain: project context used
- Roles applied:
product-managerrole,ml-engineerrole,sre-engineerrole - Data: source, volume, time range
- Key findings: top 3 insights
- Recommendations: summary of proposed changes
- Feature plan: link to generated plan or status
- Next step:
feature-devskill per work package
Integration
- follow-up skills:
analyze-prodskill (production data access, Steps 6a–6b) - Followed by:
planskill (Step 9), thenfeature-devskill (implementation) - Roles:
product-managerrole (task formulation),ml-engineerrole (analysis, modeling),sre-engineerrole (data extraction) - Skills:
test-strategyskill (validation plan),context-engineeringskill (context pipeline design, RAG, memory engineering, agent harness — for LLM/prompt recommendations),prompt-engineeringskill (technique selection, eval strategy)
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

