iso27001-annex-a-controls - SKILL.md Agent Skill

name: iso27001-annex-a-controls description: Implements ISO/IEC 27001:2022 Annex A control assessment—93 controls across Organizational, People, Physical, and Technological themes—covering Statement of Applicability (SoA), risk treatment, and implementation evidence. Trigger when building or auditing an ISMS, mapping Annex A to existing controls, preparing ISO 27001 certification, or assessing agent/MCP systems against ISO 27002:2022 guidance. Do not use for SOC 2 TSC mapping alone (use soc2-trust-services-criteria), IAM-only reviews (use access-control-identity-audit), or vendor SOC report review without ISO scope (use vendor-third-party-risk).

ISO 27001 Annex A Controls

Overview

This skill performs ISO/IEC 27001:2022 Annex A control assessments aligned with ISO/IEC 27002:2022 implementation guidance. Annex A contains 93 controls organized in four themes:

Theme	Control count	Examples
Organizational (A.5)	37	Policies, asset management, supplier relationships, incident management
People (A.6)	8	Screening, terms of employment, awareness, remote working
Physical (A.7)	14	Secure areas, equipment, clear desk, cabling security
Technological (A.8)	34	Access control, cryptography, logging, secure development, cloud services

ISO 27001 Clause 6.1.3 requires a Statement of Applicability (SoA) documenting which Annex A controls apply, justification for exclusions, and implementation status. Clause 8.1 requires operational control implementation with evidence.

This skill cross-maps to existing repository skills:

A.5.15–A.5.18, A.8.2–A.8.5 → access-control-identity-audit (IAM, privileged access, authentication)
A.8.15, A.8.16 → audit-logging-integrity (logging, monitoring, SIEM)
A.5.19–A.5.23 → vendor-third-party-risk (supplier and cloud service security)
A.8.24, A.8.9 → pci-dss-encryption-key-management (cryptography, configuration management)
A.8.25–A.8.28 → mcp-compliance-integration (secure development, application security for agents)

When to Use

Use this skill when:

Building or auditing an ISMS for ISO 27001:2022 certification or surveillance audit
Drafting or validating the Statement of Applicability (SoA)
Performing risk treatment and mapping controls to identified risks (Clause 6.1.2)
Gap assessing agent/MCP infrastructure against Annex A Technological controls
Harmonizing ISO 27001 with SOC 2, HIPAA, or PCI programs already in this repository
Collecting implementation evidence for certification body or internal audit

Do not use this skill when:

SOC 2-only TSC evidence binders without ISO scope (use soc2-evidence-collection)
PCI CDE network segmentation (use pci-dss-network-segmentation)
California consumer privacy rights (use ccpa-cpra-privacy-rights)
NIST CSF function-based gap analysis without ISO SoA deliverable (use nist-csf-2-assessment)

Core Process

Execute steps in order.

Step 1: ISMS scope and context (Clauses 4.1–4.3)

Document ISMS scope boundary: legal entities, locations, systems, processes, data types.
Identify interested parties and their requirements (customers, regulators, certification body).
For AI/agent systems, explicitly include:
- Compliance agent runtime (agent.py)
- MCP servers (Playwright, Postgres, Presidio, Terraform)
- LLM vendor subprocessors processing redacted or raw data
Artifact: isms-scope-{id}.json with UTC timestamp and approver.

Step 2: Risk assessment (Clause 6.1.2)

Identify information security risks within scope using asset-threat-vulnerability methodology.
Rate likelihood and impact; assign risk owners.
Select treatment options: mitigate, transfer, avoid, accept (with documented acceptance criteria).
Map each treated risk to one or more Annex A controls as treatment actions.
Artifact: risk-register-{id}.csv with risk ID, treatment, linked control IDs (e.g., A.8.5).

Step 3: Statement of Applicability (SoA) — Clause 6.1.3

List all 93 Annex A controls from ISO/IEC 27001:2022 Annex A.
For each control, record:

Field	Content
Control ID	e.g., A.5.15
Applicable?	Yes / No
Justification if excluded	Required for "No"
Implementation status	Implemented / Partial / Planned / N/A
Implementation reference	Policy ID, procedure, technical control
Evidence location	File path, system, audit ID

No control may be excluded without justification—certification bodies reject blank SoA rows.
For agent systems, typically mandatory controls include A.5.9 (asset inventory), A.5.23 (cloud services), A.8.2 (privileged access), A.8.15 (logging), A.8.28 (secure coding).
Artifact: statement-of-applicability-{id}.yaml.

Step 4: Theme-by-theme control assessment

Assess each theme with evidence requirements per ISO/IEC 27002:2022:

Organizational (A.5):

A.5.1: Information security policy approved by top management
A.5.2: Roles and responsibilities documented (include agent/MCP owners)
A.5.7: Threat intelligence process
A.5.24–A.5.28: Incident management (cross-ref breach-incident-response)
A.5.29–A.5.30: Business continuity

People (A.6):

A.6.3: Security awareness including AI/agent data handling
A.6.7: Remote working controls for agents accessing production evidence

Physical (A.7):

A.7.4: Physical security monitoring (data center, office)
A.7.10: Storage media (evidence drives, backup tapes)

Technological (A.8):

Delegate deep dives to cross-cutting skills (see Overview mapping)
A.8.11: Data masking (cross-ref hipaa-phi-redaction-pipeline)
A.8.12: DLP (Presidio gate in redaction.py)
A.8.23: Web filtering (MCP URL allowlists)

Document PASS/FAIL/PARTIAL per control with observation and evidence hash.

Step 5: Cross-framework mapping

Build traceability matrix linking Annex A to in-repo frameworks:

Annex A	SOC 2	HIPAA	PCI-DSS
A.8.2	CC6.1	§164.312(a)	Req 7
A.8.15	CC7.2	§164.312(b)	Req 10
A.5.19	CC9.2	BAA	Req 12.8
A.8.24	CC6.1	§164.312(a)(2)(iv)	Req 3/4

Reuse evidence from existing audits—do not duplicate collection without gap analysis.
Flag controls with no cross-map requiring ISO-specific evidence.

Step 6: MCP and agent evidence collection

Use Postgres MCP (read-only) to query asset/configuration tables for A.5.9, A.8.9 evidence.
Use Terraform MCP (mcp/terraform.mcp.json) for A.8.9 configuration baseline drift.
Use Playwright MCP only for documented URLs—capture screenshots for A.8.20 (network security) where applicable.
Never store unredacted PHI or PAN in ISO evidence binders—apply redaction.py before LLM analysis.
Hash all evidence files (SHA-256) per templates/audit-evidence-manifest.yaml.

Step 7: Internal audit and management review prep (Clauses 9.2, 9.3)

Sample 10% of applicable controls for deep evidence testing.
Prepare management review inputs: open risks, SoA changes, nonconformities, audit findings.
Document corrective actions for FAIL/PARTIAL controls with owner and due date.

Step 8: Findings and remediation tracker

Emit findings in standard format:

id: FIND-ISO-001
severity: HIGH
control_id: A.8.5
observation: "MCP service account lacks periodic access review"
recommendation: "Implement quarterly recertification per A.8.5; use access-control-identity-audit Step 7"
owner: ""
due_date: null
status: open

Prioritize: CRITICAL (certification blocker), HIGH (major nonconformity), MEDIUM (minor), LOW (observation).

Common Rationalizations

Excuse the agent might generate	Required rebuttal
"We're SOC 2 certified, so ISO 27001 is automatic."	SOC 2 TSC does not cover all 93 Annex A controls—SoA must address each control individually with ISO-specific evidence.
"We excluded all Physical controls because we're cloud-only."	Cloud-only orgs still need applicable A.7 controls for offices, laptops, and media—blanket exclusion fails certification.
"Annex A is guidance only; we don't need evidence."	ISO 27001 requires implemented controls for all applicable Annex A entries—27002 is implementation guidance, not optional.
"Agent infrastructure is out of ISMS scope."	If agents process scoped data or access production systems, they are in scope—A.8.25–A.8.28 apply.
"SoA can mark everything 'Planned' for year one."	Certification bodies expect Implemented or justified Partial with remediation plan—all-Planned SoA rows are rejected.
"Vendor SOC 2 report replaces A.5.19 assessment."	SOC report is input to supplier review, not replacement—A.5.19–A.5.22 require your organization's due diligence.

Red Flags

SoA with unexplained "No" for >20% of controls without risk-based justification
No risk register or risks not linked to Annex A treatment controls
Agent/MCP credentials absent from A.5.9 asset inventory
A.8.15 logging not covering MCP tool invocations and deanonymization events
Cloud LLM vendor missing from A.5.19 supplier register
Evidence binder contains raw ePHI or PAN without redaction
Internal audit sampled only Organizational theme, skipped Technological (A.8)
Management review not conducted within 12 months of certification cycle

Verification

ISMS scope documented including agent/MCP and LLM subprocessors
Risk register complete with treatment actions mapped to Annex A controls
Statement of Applicability covers all 93 controls with justification for exclusions
Each applicable control assessed PASS/FAIL/PARTIAL with evidence reference and SHA-256 hash
Cross-framework mapping matrix completed against SOC 2, HIPAA, PCI where in scope
Technological controls (A.8) assessed using access-control, audit-logging, vendor skills
MCP evidence collected via read-only tools without cleartext PHI/PAN in artifacts
Internal audit sample (≥10% applicable controls) executed with findings logged
Remediation tracker populated for FAIL/PARTIAL with owners and due dates
Management review inputs prepared per Clause 9.3