rcsd-data-analyst

name: RCSD Data Analyst description: This skill should be used when the user asks about "Redwood City schools", "RCSD", "school hours", "school enrollment", "school calendar", "is there school today", "next board meeting", "what's for lunch", "lunch menu", "report an absence", "IEP data", "special education", "EL percentage", "LTEL", "long-term English learner", "chronic absenteeism", "teacher diversity", "staff demographics", "teacher experience", "pupil-teacher ratio", "school site council", "SSC", "SPSA", "free and reduced lunch", "PTO", "Konstella", "ParentSquare", "which school", "board meeting", "SARC", "test scores", "CAASPP", "school budget", "RCEF", "Measure U", "expenditures", "district property", "former school site", "who leases the old campus", "watch board meeting", "compare schools", "school demographics", "meeting transcript", "board discussion", "who is my trustee", "who represents my area", "board member", "trustee area", "board president", "who is the superintendent", "new superintendent", "district cabinet", "CBO", "chief business official", "find the resolution", "board resolution", "employment contract", "find the agreement", "MOU", "board packet", "agenda item document", "warrant register", "change order", or any question about Redwood City School District schools, demographics, calendars, meetings, lunch menus, funding, staffing, trustees, board documents, or parent resources. Also activates when the user mentions a child's name in the context of school. version: 0.5.0

RCSD Data Analyst

Answer any question about Redwood City School District (RCSD) — a TK-8 public school district in Redwood City, California serving ~6,500 students across 12 schools — by reading and reasoning over local data files directly.

Core Approach

All structured data lives in data/ within the rcsd.info project (typically /Users/dew/dev/rcsd/rcsd.info/data/). The files are small JSON (under 500KB total) — read them directly with the Read tool and reason over the contents. This enables arbitrary queries, comparisons, and cross-file analysis that no pre-built tool can match.

For questions requiring live external data (lunch menus), use the bundled scripts.

Family Configuration

Check for ~/.claude/rcsd-info.local.md in the user's home directory. This stores family context for resolving child-specific questions ("What's Max having for lunch?").

Expected format:

---
children:
  - name: Max
    grade: 5
    school: orion
    program: Mandarin Immersion
  - name: Cyrus
    grade: 3
    school: orion
    program: Mandarin Immersion
---

When a user mentions a child by name, resolve their school and grade from this config. If the config doesn't exist and the user asks a child-specific question, ask which school and offer to save the config.

Data File Inventory

Read these files from data/ to answer questions. For field-by-field documentation, consult references/data-schema.md.

Schools & District

Demographics & Academics

Board Meetings (190 meetings, Aug 2020 - present)

Board Policies (619 policies, bylaws, and regulations)

School Slugs

Query Strategy

Simple lookups (single file)

Read the relevant file and extract the answer. Examples: school phone number, bell schedule, calendar check, meeting summary.

Cross-file analysis (join reasoning)

Read multiple files and reason across them. Examples: "Which schools have high EL% but low math scores?" requires joining sarc/sarc-summary.json (demographics + CAASPP) with schools.json (enrollment context).

Finding a specific named board document (resolution, contract, agreement, MOU, change order, warrant register)

Don't conclude "it isn't in our data" from one empty grep. Follow this order:

1. grep data/agenda-attachments.json by title keyword (e.g. rubalcaba, employment contract, Resolution No). It is keyed by meeting date and lists every attachment as {aid, title, url, page} — the most complete source.
2. Also grep data/document-index.json (classified) if you want to filter by type/school/year — but remember it omits unclassified items, so a miss here is not authoritative.
3. Build the public PDF URL from the attachment's meeting date + filename (in document-index.json): https://data.rcsd.info/board-packets/{meetingDate}/{filename}. (agenda-attachments.json carries the original Simbli Attachment.aspx?AID=… link; the friendly R2 mirror is board-packets/{date}/{filename} or board-packets/{aid}.pdf.)
4. If you know the meeting but not the item, read that meeting in meetings-data.json / board-memos/{date}.json to get the agenda item label, then look up its attachments.

Example: the superintendent's employment contract → agenda-attachments.json under 2026-01-21 (agenda item 12.3) → https://data.rcsd.info/board-packets/2026-01-21/Superintendent-s-Employment-Contract_Redwood-City-SD-Dr.-Christian-Rubalcaba-202.pdf.

Temporal/topical analysis (meetings corpus)

For "what has the board discussed about X?", search meetings-data.json for topic keywords in the topics array and item titles, then read meeting-summaries.json for context. For deeper detail, read the specific board-memos/{date}.json files. See references/meetings-guide.md for navigating the meeting corpus.

Board Policies queries

For "what is the district's policy on X?", follow this 3-step strategy:

1. Search Index: First, read data/policies-index.json. Perform a text or regex search on the title or code fields of the policies array to identify candidate policy codes and types (e.g. 5141.22 BP or 9223 AR).
2. Read Detail File: Once you identify the relevant policy code and type, read its individual detail JSON file directly from data/board-policies/{code}-{type}.json (e.g. data/board-policies/9223-AR.json). For a Spanish answer, the machine-translated body is in data/board-policies-es/{code}-{type}.json (contentTextEs); note it is unofficial — the English version is authoritative.
3. Analyze & Present: Extract the contentText for the core policy rules, the footnotes for statutory citations (like the CA Education or Government Code), and the crossRefs for related governance rules to synthesize an authoritative answer.

Comparative queries

Read the relevant data for all schools and present side-by-side. The data is small enough to load entirely.

For detailed cross-file query examples, consult references/query-patterns.md.

Live Data (Scripts)

Only lunch menus require live API calls. Everything else is answered from local JSON.

Lunch Menus

Fetch live from the HealthePro API using the bundled script:

node ${SKILL_DIR}/scripts/lunch-menu.mjs <slug> [date]
node ${SKILL_DIR}/scripts/lunch-menu.mjs orion tomorrow
node ${SKILL_DIR}/scripts/lunch-menu.mjs orion 2026-04-03

Date accepts: YYYY-MM-DD, today, tomorrow (defaults to today).

If the script is unavailable, call the HealthePro API directly. See references/data-schema.md for endpoint details and school-to-menuId mapping.

School Lookup (Convenience)

For quick formatted school profiles, the bundled script is available but reading schools.json directly is preferred for flexibility:

node ${SKILL_DIR}/scripts/query-school.mjs <slug> [--sped] [--meetings]
node ${SKILL_DIR}/scripts/query-school.mjs --calendar YYYY-MM-DD
node ${SKILL_DIR}/scripts/query-school.mjs --list

Data Caveats

• Cell suppression: CDE data uses null where counts are <=10 students (privacy). State this when presenting data.
• SARC year lag: 2024-25 SARCs report 2023-24 data. Note the reporting year.
• CDE data year: CDE bulk files in data/cde/ are 2024-25 (more current than SARCs).
• LTEL at elementary: Elementary schools (TK-5) typically show 0 LTELs because students haven't been ELs long enough (6+ years). At-Risk is the relevant metric for elementary.
• Absenteeism codes: Use reporting category codes (TA=All, RH=Hispanic, SE=English Learners, SS=SED, SD=Students with Disabilities, SH=Homeless). See references/data-schema.md for full code table.
• Staff ratios: null ratio means denominator FTE < 1.0 (CDE doesn't compute ratio).
• 504 plans: Not tracked by CDE; only available from OCR CRDC (lags ~5 years).
• AI-generated content: Meeting summaries and SSC membership data (extracted from PDFs via Claude) are AI-generated and labeled as such.
• Lunch menus: Published monthly; future months may not yet be available.
• Bilingual: Calendar events have en and es fields. The site has /schools/ and /escuelas/ mirrors.

Remote Fallback

If local data files are not available (e.g., repo not cloned), all JSON is published at https://data.rcsd.info/json/. Use WebFetch as a fallback. Board meeting videos are on YouTube (links in youtube-index.json).

Additional Resources

Reference Files

• references/data-schema.md — Complete field-by-field documentation of every JSON data file, plus HealthePro API details
• references/query-patterns.md — Examples of cross-file analysis queries with step-by-step approaches
• references/meetings-guide.md — Navigating the 190-meeting board corpus: structure, timestamps, transcripts

Website

• rcsd.info — school pages at /schools/{slug}/, meetings at /meetings/, budget at /budget/
• data.rcsd.info — public JSON and artifact hosting