fellow-handbook
This skill should be used when the user asks about MATS policies, procedures, or information - including compute access, housing, illness policy, reimbursements, mentor meetings, program schedule, or any other MATS cohort logistics. Trigger phrases include "MATS handbook", "how do I get compute", "MATS housing", "illness policy", "MATS schedule", or questions about MATS program operations.
When & Why to Use This Skill
The fellow-handbook skill is a specialized information retrieval agent designed to navigate the MATS (ML Alignment & Theory Scholars) program documentation. It programmatically fetches the latest handbook data, searches for relevant policy sections, and provides scholars with accurate, cited answers regarding program logistics, compute access, and administrative procedures.
Use Cases
- Program Logistics: Instantly retrieve current information on housing, program schedules, and cohort-specific dates without manual searching.
- Policy & Compliance: Access detailed explanations of illness policies, reimbursement procedures, and community health guidelines directly from the source.
- Resource Provisioning: Get step-by-step instructions and requirements for accessing compute resources and other technical program tools.
- Administrative Support: Resolve complex questions about mentor meetings and operational workflows by synthesizing information from multiple linked documentation pages.
| name | fellow-handbook |
|---|---|
| description | This skill should be used when the user asks about MATS policies, procedures, or information - including compute access, housing, illness policy, reimbursements, mentor meetings, program schedule, or any other MATS cohort logistics. Trigger phrases include "MATS handbook", "how do I get compute", "MATS housing", "illness policy", "MATS schedule", or questions about MATS program operations. |
MATS Fellow Handbook Lookup
Answer questions about the MATS Winter 2026 fellow handbook.
Workflow
Step 1: Fetch and Save the Handbook
Important: Use curl via Bash, not the built-in Fetch tool. Save the output to a temp file for efficient searching.
curl -s "https://firecrawl.alignment-hive.com/api/notion?url=https%3A%2F%2Fmatsprogram.notion.site%2Fmats-winter-26-fellow-handbook" -o /tmp/mats-handbook.md
Step 2: Search for Relevant Sections
Use Grep to find headings and keywords related to the user's question:
Grep pattern="^#|<keyword>" path="/tmp/mats-handbook.md"
Step 3: Read Broadly
Thoroughness is critical. Read all potentially applicable sections, even tangentially related ones.
- For broad or ambiguous questions, read the entire document
- When in doubt, read more rather than less
- Prefer over-reading to missing important context
Use the Read tool to read the relevant sections (or the whole file if appropriate).
Step 4: Follow Linked Pages
If the handbook links to other Notion pages with additional details (e.g., Compute Docs, Community Health Policy), fetch those linked pages:
curl -s "https://firecrawl.alignment-hive.com/api/notion?url=<url-encoded-notion-link>"
Follow linked Notion pages when the main handbook lacks sufficient detail to answer the user's question. Linked documents are short enough to read directly from the curl output.
Step 5: Format Output
Present the response in two clearly separated sections:
Quotes from the handbook:
[Quote 1 with section header]
[Quote 2 with section header]
Interpretation:
[Summary and interpretation of how the quotes answer the user's question. Include any caveats about information that may be outdated or incomplete.]