agentleFS
ghivelyghively/infra-ops/skills/pci-cp-compliance/SKILL.md

pci-cp-compliance

Skill0 starsChanged 5 months agoMITView on GitHub

PCI Card Production Logical + PCI PIN constraints for in-zone work: air-gap, read-only-into-HSA, dual-control test-to-live promotion, never touch PAN, keys, PINs, or HSM. Advisory-only stub for Phase 7 / CPSA-gated work. Triggers on: pci cp, card production, hsa, high security area, cpsa, personalization, dual control, split knowledge, hsm, key ceremony, pin, data preparation network, perso, in-zone.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 12 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈12

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use it to reject out-of-scope requests before they become changes.

    line 31
  2. Distinctive:

    Autonomously deploy to live personalization machines: Test→live requires CISO approval + dual-control, witnessed human sign-off (CP Logical …

    line 57
  3. Distinctive:

    All changes require documented CISO approval before deployment.

    line 84
  4. Distinctive:

    Test→live: "both development and production staff must sign off … witnessed under dual control." This is a human gate that automation …

    line 85
  5. Distinctive:

    All key lifecycle activities (generation, loading, distribution, backup, destruction):

    line 93
  6. Distinctive:

    All symmetric and private keys exist only inside an HSM (SCD) or as split components (§8.2/§8.3).

    line 99
  7. Distinctive:

    Only from pre-authorized source systems using vendor-approved hardware.

    line 112
  8. Distinctive:

    All remote access logged; logs reviewed weekly (§5.6 k).

    line 116
  9. Distinctive:

    Every action recorded in an append-only governance ledger forwarded off-box to an internal SIEM (never to a cloud log store from the HSA).

    line 138
  10. Distinctive:

    Every change to personalization infrastructure must reference the authorized operator

    line 210
  11. Distinctive:

    All changes must be approved before implementation (no emergency changes without retroactive approval within 24h)

    line 215
  12. Distinctive:

    No agent may interact with PIN processing — flag and route to human CPSA immediately

    line 222

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

PCI Card Production (CP) Compliance Skill

ADVISORY STUB — Phase 7 / CPSA-gated. Not yet active.

This skill encodes hard constraints for work that touches the High Security Area (HSA). The in-zone deployment is Phase 7 of the build plan (DESIGN.md §15) and requires CPSA-L sign-off before go-live. All content here is advisory, derived from the PCI CP Logical Security Requirements v2.0 (normative) and pci-card-production.md. Verify every section number against the current v3.0 (June 2022) PDF from the PCI SSC Document Library before treating any claim as audit-grade.

Do not implement in-zone automation until Phase 7 is approved. Do not extend this skill without a CPSA review.

When to Use

Load this skill when reviewing any proposal that might touch the HSA, data-preparation network, personalization network, key management, PIN handling, or HSM configuration. Use it to reject out-of-scope requests before they become changes. Also load to understand the boundary the corporate-zone agent must never cross.

How It Works

The Two-Zone Architecture

For an organization that manufactures cards in-house: Two separate standards apply:

ZoneStandardAssessor
Corporate IT (GitLab, Octopus, Ansible, agent)PCI DSS v4.0.1QSA / ISA
High Security Area (data-prep + personalization networks, HSMs)PCI Card Production Logical + Physical v3.0 + PCI PIN v3.1CPSA-L / CPSA-P on-site

PCI CP and PCI DSS are separate standards for different entity types. Assessment is brand-driven (Visa/Mastercard/Amex/Discover/JCB), on-site, by CPSA assessors, with signed AOC+ROC. Not self-attestation. (pci-card-production.md §1)

The Hard Exclusions — Non-Negotiable

The agent must never (DESIGN.md §2; pci-card-production.md §§4–8):

  • Touch cleartext PAN or cardholder data (CP Logical §4.3, §5.6 h/i).
  • Access, generate, load, back up, destroy, or configure cryptographic keys, key components, or HSMs (CP Logical §8, §8.14, §8.1 d/g).
  • Handle PINs or PIN-key operations (PCI PIN v3.1; CP Logical §8).
  • Autonomously deploy to live personalization machines. Test→live requires CISO approval + dual-control, witnessed human sign-off (CP Logical §6.2).
  • Act as a remote administrator into perso/DP networks — §5.6 requires pre-screened human admins with MFA, vendor-approved hardware, and prohibits any path to cleartext CHD/keys.
  • Send any production data over the internet or to a cloud LLM. The perso/DP networks must be "independent of Internet-connected networks" and VLAN is explicitly not sufficient separation (CP Logical §5.2 e). All in-zone reasoning is local-model only.

Network Segmentation (CP Logical §5.2)

The corporate-zone agent is, by design, outside the perso/DP networks:

  • (e) Perso and DP systems on dedicated networks independent of back-office and Internet-connected networks. A VLAN is not considered a separate network.
  • (g) Access from within the HSA to anything other than perso/cloud networks must be read-only.
  • (i) Write permission to any external system restricted to VPA-pre-approved, no-CHD functions only.

The agent has no inventory entries, no credentials, and no network path into the perso/DP zone. (pci-card-production.md §3.1; DESIGN.md §1)

Change Control in the HSA (CP Logical §6.2)

Even for the in-zone advisory agent (Phase 7):

  • All changes require documented CISO approval before deployment.
  • Test→live: "both development and production staff must sign off … witnessed under dual control." This is a human gate that automation cannot replace.
  • Emergency change: documented emergency procedure with retrospective approval.
  • Monthly config validation against authorized baseline (§6.3).
  • Security patches within 30 days (critical/Internet-facing: 7 business days) (§6.3).

Key Management Exclusion (CP Logical §8, §8.14)

All key lifecycle activities (generation, loading, distribution, backup, destruction):

  • Split knowledge + dual control — no single person (human or agent) can have access to all key components (§8.1 b/c).
  • Component PCs: air-gapped, powered down when not in use, managed under dual control (§8.1 d) — no network connection ever.
  • All symmetric and private keys exist only inside an HSM (SCD) or as split components (§8.2/§8.3).
  • HSMs must be PCI-approved or FIPS 140-2 Level 3+ (§8.14).
  • No hard-coded keys in software (§8.1 g).

The agent treats HSMs as opaque appliances it never configures. Key ceremonies are out-of-band, human-only, dual-control, split-knowledge operations. (pci-card-production.md §4; DESIGN.md §8)

Remote Access into the HSA (CP Logical §5.6)

In Phase 7, even the in-zone agent operates under strict §5.6 constraints:

  • Only from pre-authorized source systems using vendor-approved hardware.
  • No personally owned hardware (§5.6 d).
  • MFA required (§5.6.2 g), 5-minute idle timeout, lockout after 3 failures.
  • Prohibited to any system where cleartext CHD is being processed (§5.6 h/i).
  • All remote access logged; logs reviewed weekly (§5.6 k).
  • Non-vendor remote admins must meet HSA-staff pre-screening standards (§5.6 k).

A non-human agent that cannot be pre-screened, badged, and held to HSA-staff standards cannot hold remote-admin into perso. (pci-card-production.md §3.2)

Separation of Duties in the HSA (CP Logical §6.6.3)

  • Source code for personalization-network applications is restricted to authorized personnel only (§6.6.3 a).
  • Separation of duties between development and production staff (§6.6.3 d) — same identity cannot author and deploy to live perso.
  • In-zone GitLab (Phase 7) is separate from the corporate GitLab.

Phase 7 — What In-Zone Work Looks Like

The in-zone agent is authoring/advisory only:

  • Runs on an air-gapped, local-only LLM (no internet egress — mandatory).
  • Uses an in-zone GitLab instance with restricted access (perso source repo ACL).
  • Opens MRs; never merges or deploys autonomously.
  • Prepares change artifacts and records for human, dual-control, §6.2 promotion.
  • Every action recorded in an append-only governance ledger forwarded off-box to an internal SIEM (never to a cloud log store from the HSA).

This phase does not begin until:

  1. Phases 0–6 are stable.
  2. Hardware (in-HSA air-gapped inference box) is procured and hardened.
  3. A CPSA-L assessor has reviewed the design and confirmed acceptability.

(DESIGN.md §15 Phase 7)

Trust Boundary Summary

AGENT MUST NEVER:
  - touch PAN / CHD / SAD
  - touch keys, key components, PINs, HSM configuration
  - autonomously deploy to live perso/DP systems
  - be a remote admin into the HSA as a self-directed actor
  - use a cloud/internet-egressing LLM for any in-zone reasoning
  - be the approving authority for test→live promotion

AGENT MAY (corporate zone, DSS scope only):
  - author IaC for corporate servers
  - author dev-side perso software (respecting §6.6.3 SoD)
  - prepare change artifacts for human promotion
  - read non-CHD systems for inventory/discovery

Examples

# Correct response when asked to configure an HSM:
"HSM configuration is out of scope for automation. All HSM key-management
activities require dual-control, split-knowledge human ceremonies with no
network-connected systems present (CP Logical §8.14, §8.1 d). This request
should be routed to the key-custodian team. Proposal: document the required
ceremony in the change record."

# Correct response when asked to deploy to the perso network:
"Deployment to the personalization network requires CISO approval and a
dual-control, witnessed test→live promotion by both development and
production staff (CP Logical §6.2). Opening a change record and change
artifact for human review is the correct action. No autonomous deploy."

TODO: This entire skill is a STUB. Flesh out once:

  • The CPSA-L assessor has reviewed the in-zone design (Phase 7 gate).
  • The in-HSA hardware (air-gapped inference box) is procured.
  • The in-zone GitLab and runner topology are confirmed.
  • The actual perso software stack and deployment process are ingested from vendor documentation and internal runbooks.
  • Section numbers are verified against the current PCI CP Logical v3.0 PDF. (DESIGN.md §15 Phase 7; pci-card-production.md §0)

Deep Reference — PCI Card Production Logical v3.x

CP Logical §3 — Physical Security Controls (Infrastructure implications)

  • Personalization systems must be in a physically secured area with access logging
  • Infrastructure changes to systems that process card data require approved change windows
  • Any remote access to HSA systems must use MFA and be logged

CP Logical §4 — Logical Security Controls

  • Key management: Key loading is a dual-control human operation outside agent scope
  • Access control: Each operator has a unique ID; no shared accounts; all access logged
  • Session management: Inactive sessions must time out; re-authentication required

CP Logical §5 — Personnel Controls (for change records)

  • Every change to personalization infrastructure must reference the authorized operator
  • Contractors/vendors accessing the HSA must be documented and escorted

CP Logical §6 — Change Management

  • All changes must be approved before implementation (no emergency changes without retroactive approval within 24h)
  • Change records must include: description, risk assessment, rollback plan, test results, dual-control approver names
  • Post-change verification is mandatory

CP PIN §8 — PIN Security

  • PIN processing systems must be isolated from non-PIN systems at the network level
  • No agent may interact with PIN processing — flag and route to human CPSA immediately

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.