| name | knowledge-base-onboarding |
| description | Onboards a new Robotics or AI research area, project folder, or lab knowledge base into a structured source inventory, concept map, artifact map, and next-action plan. Use when the user starts a new area, inherits notes/code/papers, or needs a reusable research memory structure. |
Knowledge Base Onboarding
Use this before deep synthesis when the user has scattered papers, notes, code, links, or project folders.
Inputs
- Research topic, project folder, paper list, notes directory, Zotero/BibTeX export, repository, or dataset links.
- User goal: field onboarding, project takeover, survey, paper writing, experiment reproduction, or code extension.
- Constraints: time budget, venue, course requirement, available tools, and private data boundaries.
Read references/onboarding-map.md when building a multi-source onboarding artifact.
Workflow
- Identify the user's immediate decision: what they need to understand, write, build, or debug.
- Inventory available sources without assuming they are complete or trustworthy.
- Classify sources into papers, code, datasets, experiments, notes, figures, decisions, and open questions.
- Build a concept map: key terms, methods, benchmarks, claims, and recurring failure modes.
- Build an artifact map: where evidence lives and what is missing.
- Mark provenance and trust level for each source.
- Route follow-up work to
paper-search-protocol, paper-reading-card, evidence-matrix-builder, related-work-positioning, or coding skills.
Output
## Knowledge Base Onboarding Map
## Source Inventory
| Source | Type | Location | Trust | Next action |
| --- | --- | --- | --- | --- |
## Concept Map
## Artifact Map
## Missing Context
## Recommended Next Skills
Rules
- Do not summarize a source as read if only its title, README, or abstract was inspected.
- Do not mix verified facts with the user's hypotheses.
- Do not copy private local paths into public artifacts.
- Do not build a taxonomy before the source inventory is explicit.
- Do not treat a knowledge base as complete without naming coverage gaps.
Validation
- Every source has type, location, and trust status.
- The concept map separates definitions, methods, benchmarks, and claims.
- Missing sources and stale notes are visible.
- The next action is a concrete skill route, not a vague "learn more".
Completion
Done means the user has a navigable map of what exists, what is reliable, what is missing, and which skill should run next.