agentsclimarketplace

Vendor third party risk

Skill vaquarkhan/compliance-agent-skills/skills/vendor-third-party-risk

30 Agent Skills for deterministic USA compliance auditing of AI agents — HIPAA, NIST AI RMF, FERPA, COPPA, PCI-DSS, SOC 2, FedRAMP, CMMC, GLBA, privacy & GDPR. Presidio PHI redaction, MCP templates, Pydantic AI, VS Code & JetBrains plugins.

Install
npx -y skills add vaquarkhan/compliance-agent-skills --skill vendor-third-party-risk

Assembled 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.

What its author says it does

Copied from the file, not written here

Performs third-party and vendor risk assessments for compliance programs—security questionnaires, SOC 2 report review, control inheritance, and ongoing monitoring—for LLM, cloud, MCP, and GRC tool vendors (Vanta/Drata alternative operational mindset). Trigger when onboarding vendors, annual vendor reviews, or assessing subprocessor risk for HIPAA, PCI, and SOC 2. Do not use for BAA clause legal analysis (use hipaa-baa-vendor-assessment) or MCP technical hardening (use mcp-compliance-integration).

SKILL.md

7.6 KB, as published. Nobody here has run it

Vendor Third-Party Risk

Overview

This skill implements third-party risk management (TPRM) for compliance programs—operating with the rigor of GRC platforms (Vanta, Drata, Secureframe) but without assuming a specific vendor tool. It complements hipaa-baa-vendor-assessment (legal BAA focus) with operational security assessment:

  • Inherent/residual risk scoring
  • SOC 2 Type II report review (bridge letters, subservice orgs)
  • Security questionnaire mapping (SIG Lite, CAIQ, custom)
  • Continuous monitoring of vendor status (breach news, certification expiry)
  • Control inheritance decisions for SOC 2 carve-out reports

Applies to LLM providers, cloud hosts, MCP operators, observability SaaS, and identity providers.

When to Use

Use this skill when:

  • Onboarding a new vendor processing customer data, ePHI, or CHD
  • Annual vendor recertification in compliance calendar
  • Reviewing vendor SOC 2 report for subservice organization reliance
  • Completing security questionnaires sent to/from vendors
  • Assessing LLM/AI vendor risk beyond BAA (model security, logging, subprocessors)
  • Building vendor risk register for SOC 2 CC9.2 and HIPAA §164.308(b)
  • Evaluating GRC automation tools themselves as vendors

Do not use this skill when:

  • Drafting BAA legal clause analysis (use hipaa-baa-vendor-assessment)
  • MCP OAuth configuration (use mcp-compliance-integration)
  • Internal employee access reviews (use access-control-identity-audit)

Core Process

Execute steps in order.

Step 1: Vendor tiering

  1. Classify vendors by inherent risk:
TierCriteriaReview depthExamples
T1 CriticalePHI/CHD processing, production accessFull assessment + SOC 2 + annual onsite/virtualLLM with BAA, cloud prod, MCP host
T2 HighConfidential data, internal network accessSOC 2 + questionnaireSIEM, IdP, ticketing
T3 MediumLimited data, no prod accessQuestionnaire + certification checkMarketing SaaS
T4 LowNo sensitive dataLightweight attestationOffice supplies
  1. Document tier in vendor risk register.

Step 2: Due diligence data collection

For each in-scope vendor, collect:

  1. SOC 2 Type II report (≤12 months old) + bridge letter if gap period
  2. ISO 27001/SOC 1 if relevant to service type
  3. Completed security questionnaire (map to TSC controls)
  4. Pen test summary or redacted executive summary
  5. Subprocessor list and change notification process
  6. Insurance (cyber liability) certificates where required
  7. For LLM vendors: data retention, training policy, region, incident history

Store with SHA-256 hashes; redact customer references if sharing internally.

Step 3: SOC 2 report review methodology

  1. Verify report scope covers services in use (not just corporate IT).
  2. Review exceptions and management responses—unresolved exceptions = gap.
  3. Map subservice organizations (carve-out vs inclusive):
    • Carve-out: obtain subservice SOC 2 and map to your CC9.2 monitoring
  4. Check period of time covers full examination window needed.
  5. Document complementary user entity controls (CUECs) you must implement.

Step 4: Control mapping and gaps

  1. Map vendor controls to your TSC/HIPAA/PCI requirements:
Your control needVendor evidenceGap?
CC6.1 MFASOC 2 CC6.1 no exceptionsNo
HIPAA encryptionBAA + SOC C1Verify region
PCI segmentationNot in vendor scopeYou implement
  1. Gaps become compensating controls or remediation conditions before production use.

Step 5: Risk scoring

  1. Calculate inherent risk (data sensitivity × access level × criticality).
  2. Calculate residual risk after vendor controls and your CUECs.
  3. Score 1–5 or Low/Med/High; T1 vendors with High residual risk block production until remediated.
  4. Privacy officer / security committee approval for T1 High residual.

Step 6: Contractual and operational requirements

  1. Ensure contracts include:
    • Right to audit / SOC 2 delivery annually
    • Breach notification SLA (≤24–72 hours to you)
    • Subprocessor notification and objection rights
    • Data return/destruction at termination
  2. For LLM: no training on customer data, zero retention option documented.
  3. Cross-reference hipaa-baa-vendor-assessment for BAA execution.

Step 7: Ongoing monitoring (continuous TPRM)

  1. Quarterly for T1: recheck SOC 2 expiry, subprocessor changes, breach news (RSS, CISA).
  2. Annually for T2–T3: full reassessment refresh.
  3. Integrate with soc2-ccm-continuous-monitoring:
    • Alert on vendor certification expiry 90 days out
    • Auto-create reassessment ticket
  4. Off-boarding: verify access revoked, data deleted, tokens rotated.

Step 8: Risk register and reporting

  1. Maintain vendor risk register: tier, score, last review, next review, open gaps.
  2. Executive dashboard: T1 vendor health, overdue reviews, open findings.
  3. Export evidence for SOC 2 CC9.2 and auditor requests.

Common Rationalizations

Excuse the agent might generateRequired rebuttal
"Big vendor is safe—skip questionnaire."T1 vendors require evidence-based assessment regardless of brand—SOC 2 exceptions happen at scale vendors too.
"SOC 2 report from 2 years ago is fine."Reports >12 months without bridge letter fail freshness standards for T1 vendors.
"Carve-out means we ignore subservice risk."Carve-out shifts monitoring obligation to you (CC9.2)—obtain and review subservice SOC 2.
"Questionnaire completed by sales engineer is sufficient."Security assessments require validated responses mapped to controls—not marketing answers.
"We use Vanta/Drata—they manage vendors."GRC tools orchestrate workflows—you still own risk decisions and evidence quality.
"LLM vendor passed our pen test."Your vendor pen test ≠ their production security—require their SOC 2 and subprocessors review.

Red Flags

  • T1 vendor in production without current SOC 2 or equivalent
  • Unresolved SOC 2 exceptions for controls you rely on
  • Subprocessor added with ePHI access and no review within 90 days
  • Vendor breach in news with no impact assessment documented
  • Residual risk High with no committee approval
  • Vendor access still active after contract termination
  • Security questionnaire answers contradict SOC 2 report scope

Verification

  • All vendors tiered T1–T4 in risk register
  • T1/T2 due diligence package complete (SOC 2, questionnaire, subprocessors)
  • SOC 2 reports reviewed for scope, exceptions, period, and CUECs
  • Control gaps documented with compensating controls or blockers
  • Inherent and residual risk scores calculated with approval for high residual T1
  • Contractual security clauses verified (audit rights, breach SLAs, termination)
  • BAA cross-reference completed for ePHI vendors
  • Ongoing monitoring schedule configured with expiry alerts
  • Off-boarding checklist defined and tested for sample vendor
  • Risk register exported with evidence hashes for SOC 2 CC9.2

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.