Ferpa education records
Skill vaquarkhan/compliance-agent-skills/skills/ferpa-education-records
Implements FERPA (20 U.S.C. §1232g; 34 CFR Part 99) protections for student education records in EdTech, LMS integrations, and AI tutoring agents—school official exceptions, legitimate educational interest, directory information, parent/eligible student rights, and vendor DPAs. Trigger when K-12 or higher-ed systems process student records, deploying AI agents in classrooms, auditing EdTech subprocessors, or assessing LLM use with FERPA-regulated data. Do not use for HIPAA PHI in clinical settings (use hipaa-technical-safeguards), COPPA under-13 consumer apps without school context (use coppa-children-privacy), or general PII without education record nexus (use us-state-privacy-laws).From its SKILL.md
npx -y skills add vaquarkhan/compliance-agent-skills --skill ferpa-education-recordsAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
7.8 KB, ~1.7k tokens by cl100k_base, as published. Nobody here has run it
FERPA Education Records
Overview
The Family Educational Rights and Privacy Act (FERPA) protects education records—records directly related to a student maintained by an educational agency or institution (or party acting on its behalf).
| Concept | Regulation | Agent/EdTech relevance |
|---|---|---|
| Education record | 34 CFR §99.3 | Grades, discipline, IDs, biometrics, AI chat logs tied to student |
| Personally identifiable information (PII) | §99.3 | Name + student ID in agent prompts = FERPA PII |
| School official | §99.31(a)(1) | LLM vendor may be school official with DPA + LEA |
| Legitimate educational interest | §99.31(a)(1)(i) | Agent must have LEA-defined purpose |
| Directory information | §99.37 | Opt-out affects what agents may disclose |
| Parent/eligible student rights | §99.10–99.12 | Access, amend, consent to disclose |
| Recordation | §99.32 | Audit log of disclosures from agent systems |
HIPAA interaction: Most K-12 health records in education records are FERPA, not HIPAA (except school nurse as covered entity in narrow cases—legal review).
AI agents: Tutoring bots, grading assistants, and analytics pipelines that store prompts with student identity are high-risk FERPA systems.
When to Use
Use this skill when:
- EdTech SaaS or district deploys AI agents accessing student data
- Drafting Data Processing Agreements with LLM/MCP vendors as school officials
- Legitimate educational interest analysis for new agent features
- Parent consent or notification for directory information / disclosures
- Audit of agent logs, MCP Postgres, or Slack for improper redisclosure
- State student privacy laws overlay (SOPIPA, PPRA, state §)—FERPA is floor
Do not use this skill when:
- De-identified research datasets with no student IDs (may be outside FERPA—confirm)
- Pure COPPA consumer app not acting as school agent (use
coppa-children-privacy) - Employee (faculty) HR records (not education records)
- PCI payment for tuition without student academic data
Core Process
Execute steps in order.
Step 1: Education record inventory
- Map systems: LMS, SIS, agent gateway, MCP evidence DB, LLM vendor logs.
- Classify data elements per student:
- Education record vs directory information vs de-identified
- Biometric (voice/face in agent)—treat as sensitive PII
- Document who maintains each record (LEA vs vendor).
- Artifact:
templates/ferpa-school-official.yamlinventory section.
Step 2: School official and subcontractor chain
- Vendor qualifies as school official only if (34 CFR §99.31(a)(1)):
- Performs institutional service/function
- Under direct control regarding use/maintenance of records
- DPA meets §99.31(a)(1) requirements (use template)
- Map LLM provider role: subprocessors require LEA approval and flow-down terms.
- MCP servers holding student data = extensions of vendor/LEA—same control requirements.
- Prohibit training vendor models on identifiable education records without consent.
Step 3: Legitimate educational interest (LEI)
- For each agent capability, document:
- Educational purpose (tutoring, feedback, compliance audit of IT—not marketing)
- Scope of records accessed (minimum necessary)
- Role of user (teacher, admin, student)
- Deny agent access where LEI cannot be articulated (e.g., unrelated LLM experimentation).
- Align with
hipaa-privacy-minimum-necessaryconceptually—FERPA uses LEI instead.
Step 4: Consent and directory information
- Directory information (name, email, grade level)—annual notice + opt-out per §99.37.
- Agent must not combine directory info with non-directory records for unauthorized parties.
- Redisclosure rules §99.33: agent outputs to unauthorized third parties = violation.
- Parent/eligible student consent required for non-excepted disclosures (§99.30).
Step 5: Technical safeguards for agents
- Authentication: student vs teacher RBAC (
access-control-identity-audit). - Redaction: use
hipaa-phi-redaction-pipelinepatterns for student names/IDs in logs sent to LLM—prefer pseudonymous session IDs in prompts. - Audit trail §99.32: log each disclosure from agent (who, what record, recipient, purpose).
- Deanonymization in
agent.py— disable for student-facing production paths unless authorized.
Step 6: Parent and eligible student rights
- Inspection §99.10: provide access to education records including agent interaction logs tied to student.
- Amendment §99.20: process to correct inaccurate records (including wrong AI-generated summaries in record).
- Hearing procedures if amendment denied.
- DSAR-style workflows — use structured export, not raw LLM summarization of other students' data.
Step 7: Breach and improper disclosure
- FERPA has no federal breach notification statute like HIPAA. However:
- Contract and state laws may require notification
- FTC and state AG for EdTech deceptive practices
- Route confirmed leaks to
breach-incident-response+ legal. - Document in disclosure log §99.32.
Step 8: Evidence and audit
- DPA signed with all school officials (LLM, MCP host, analytics).
- LEI matrix approved by LEA privacy officer.
- Annual FERPA training for staff operating agents.
- SHA-256 evidence manifest for audit binders.
Common Rationalizations
| Excuse | Rebuttal |
|---|---|
| "We're a vendor—not subject to FERPA." | Vendors handling records on behalf of LEA are bound by contract and §99.31(a)(1) school official terms. |
| "Chat logs aren't education records." | Logs directly related to a student and maintained by institution/vendor are records under §99.3. |
| "LLM is ephemeral—no FERPA." | Vendor retention, caching, and training policies determine compliance—not user assumption. |
| "De-identified prompts are fine." | Re-identification via session context may still link to student—use pseudonymization policy. |
| "Teachers can paste any student data into the agent." | LEI and minimum access must govern—even teacher misuse is institutional control failure. |
Red Flags
- LLM vendor DPA missing school official / FERPA clauses
- Agent prompts contain student names + grades with no LEI documentation
- MCP Postgres shared across schools without tenant isolation
- No §99.32 disclosure logging for agent exports
- Directory information used in agent training data
- Parent access request answered by unredacted LLM dump of all chats
Verification
- Education record inventory and maintainer roles documented
- School official DPAs executed for LLM/MCP/analytics vendors
- LEI matrix approved for each agent capability
- Directory information notice and opt-out process current
- Technical RBAC, redaction/pseudonymization, and audit logging implemented
- Parent/eligible student access and amendment procedures defined
- Improper disclosure playbook linked to incident response
- Evidence package with privacy officer sign-off
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.
Gives 0 of the 12 instructions most education skills give in ~1.7k tokens
Counted across 169 of the 171 authors here whose files we hold, read 2026-08-07
- Search for existing resources before creating new onesin 7 of 169, across 3 files
- Check tool responses for errors before proceedingin 7 of 169, across 3 files
- Reduce request frequency on rate limit errorsin 7 of 169, across 3 files
- Confirm connection status is ACTIVE before running workflowsin 7 of 169, across 3 files
- Execute prerequisite steps first in workflowsin 7 of 169, across 3 files
- Handle pagination by fetching until exhaustedin 7 of 169, across 3 files
- Re-authenticate if the connection expiredin 6 of 169, across 2 files
- Always call RUBE_SEARCH_TOOLS first to get schemasin 6 of 169, across 2 files
- Pass strictly schema-compliant tool argumentsin 6 of 169, across 2 files
- Run the skill generator if the shared file is missingin 6 of 169, across 3 files
- Create the coursein 5 of 169, across 3 files
- List enrolled studentsin 5 of 169, across 3 files
Said here and by no other author read
- inventory maintained education records and data elements
- execute data processing agreements with all vendors
- deny agent access without articulated educational interest
- restrict vendor model training on identifiable records without consent
- apply minimum necessary access controls for each role
- redact student identifiers from logs sent to external models
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.