usap-skills / cloud-infra
CLAUDE.md4 starsChanged 2 months agoApache-2.0View on GitHub
This file is the authoritative domain guide for the cloud-infra/ skill directory in the USAP skills repository. It documents all skills in this domain, their tooling, coverage scope, inter-skill workflows, domain best practices, and MITRE ATT&CK cloud matrix coverage. The cloud-infra/ domain addresses security across five interconnected coverage areas: All five skills are read_only for discovery and analysis. Configuration changes, remediation actions, and OT network changes are mutating and require explicit human approval. OT environment changes additionally require safety review…
In this repository100 other files it gives its agents.
- agents/CLAUDE.mdCLAUDE.md
- appsec-devsecops/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- detection/CLAUDE.mdCLAUDE.md
- governance/CLAUDE.mdCLAUDE.md
- identity-access/CLAUDE.mdCLAUDE.md
- pentest/CLAUDE.mdCLAUDE.md
- platform-ai/CLAUDE.mdCLAUDE.md
- red-team/CLAUDE.mdCLAUDE.md
- response/CLAUDE.mdCLAUDE.md
- risk-compliance/CLAUDE.mdCLAUDE.md
- system-security/CLAUDE.mdCLAUDE.md
- webapp-security/CLAUDE.mdCLAUDE.md
- agent-integrity-monitorSkill
- ai-agent-securitySkill
- ai-ethics-governanceSkill
- ai-red-teamingSkill
- api-security-postureSkill
- appsec-code-reviewSkill
- appsec-customizeSkill
- attack-path-analysisSkill
- attack-surface-managementSkill
- behavioral-analyticsSkill
- build-integritySkill
- ciso-brief-generatorSkill
- cloud-security-postureSkill
- cloud-workload-protectionSkill
- compliance-mappingSkill
- container-image-scanSkill
- containment-advisorSkill
- continuous-pentestingSkill
- credential-attacksSkill
- cryptography-key-managementSkill
- cyber-insuranceSkill
- data-security-classificationSkill
- deception-honeypotSkill
- detection-engineeringSkill
- devsecops-pipelineSkill
- endpoint-os-securitySkill
- enterprise-risk-assessmentSkill
- findings-trackerSkill
- finding-triageSkill
- forensicsSkill
- guardrailSkill
- iac-securitySkill
- identity-access-riskSkill
- incident-classificationSkill
- incident-commanderSkill
- insider-physical-riskSkill
- internal-audit-assuranceSkill
- knowledge-managementSkill
- metrics-reportingSkill
- network-exposureSkill
- orchestratorSkill
- os-hardeningSkill
- ot-iot-device-securitySkill
- owasp-top10-classifierSkill
- patch-candidateSkill
- pentest-reportingSkill
- persona-coverage-auditSkill
- The first 60 shown.
The file
We split this CLAUDE.md into the rules it gives an agent
12 rules. About 0 are common, 5 are distinctive, and 7 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
7
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 17
All five skills are read_only for discovery and analysis.
- Very specific: line 45
All tools accept --help for full usage and --output json for machine-readable output.
- Distinctive: line 68
Not applicable: Tool is out of scope for that environment type
- Very specific: line 137
Link cloud posture findings to IaC policies: When the cloud-security-posture skill detects a misconfiguration that originates from a …
- Very specific: line 139
Document remediation commands; never auto-execute them: All CLI remediation commands (AWS CLI, Azure CLI, gsutil, kubectl) are included in …
- Distinctive: line 356
When adding a new skill to this domain:
- Very specific: line 358
Place the skill directory under cloud-infra/<slug>/.
- Distinctive: line 360
Update this CLAUDE.md Skills Catalog table and Cloud Provider Coverage Matrix.
- Very specific: line 361
Update the root README.md Domain Index table entry for Cloud & Infra.
- Very specific: line 362
Update domains/cloud-infra.md with the new skill slug and level.
- Distinctive: line 363
Ensure all MITRE ATT&CK technique mappings relevant to the new skill are added to the coverage section above.
- Distinctive: line 364
Add cascade rules to the Cascade and Integration Map.
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md 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.
Files
Cloud Infrastructure Security Domain — CLAUDE.md
This file is the authoritative domain guide for the cloud-infra/ skill directory in the USAP skills repository. It documents all skills in this domain, their tooling, coverage scope, inter-skill workflows, domain best practices, and MITRE ATT&CK cloud matrix coverage.
Purpose
The cloud-infra/ domain addresses security across five interconnected coverage areas:
- Cloud Security Posture (CSPM): Continuous evaluation of cloud resource configurations against CIS Benchmarks across AWS, Azure, and GCP. Detects misconfigurations, drift from known-good baselines, and maps findings to compliance frameworks (SOC 2, PCI DSS, HIPAA, NIST 800-53).
- Infrastructure-as-Code (IaC) Security: Static analysis of Terraform, CloudFormation, Pulumi, and Helm charts before deployment. Enforces security gates in CI/CD pipelines so misconfigurations are caught in code review, not in production.
- Endpoint and OS Security: Hardening assessment and EDR coverage evaluation for Windows, Linux, macOS, and containerized endpoints. Operates against CIS OS Benchmarks and DISA STIG standards.
- OT/IoT Device Security: Security posture assessment for operational technology, industrial control systems (ICS/SCADA), and IoT devices. Applies NIST SP 800-82, IEC 62443, NERC CIP, and the Purdue Enterprise Reference Architecture. Availability and safety take priority over confidentiality in this domain.
- Cloud Workload Protection (CWPP): Runtime security for containers and serverless functions. Detects container escape indicators, anomalous process behavior in pods, serverless permission sprawl, and CWPP tool coverage gaps. Complements cloud-security-posture with a runtime focus.
All five skills are read_only for discovery and analysis. Configuration changes, remediation actions, and OT network changes are mutating and require explicit human approval. OT environment changes additionally require safety review and operations director co-approval.
Skills Catalog
| Skill | Slug | Primary Tool | Coverage |
| Cloud Security Posture | cloud-infra/cloud-security-posture | cloud-security-posture_tool.py | CIS Benchmarks (AWS/Azure/GCP), CSPM, drift detection, compliance mapping |
| IaC Security | cloud-infra/iac-security | iac-security_tool.py | Terraform/CloudFormation misconfigs, Kubernetes RBAC, Helm hardening, CI/CD gate policy |
| Container Image Scan | cloud-infra/container-image-scan | container-image-scan_tool.py | Trivy/Grype/Snyk finding classification, base-image vs. app-dependency vs. implanted-layer triage, CI/CD block/fix/track/accept gate |
| Endpoint & OS Security | cloud-infra/endpoint-os-security | endpoint-os-security_tool.py | CIS OS Benchmarks, DISA STIG, EDR coverage, Windows/Linux/macOS hardening |
| OT/IoT Device Security | cloud-infra/ot-iot-device-security | ot-iot-device-security_tool.py | ICS/SCADA, NIST SP 800-82, IEC 62443, NERC CIP, Purdue Model, IoT baseline |
| Cloud Workload Protection | cloud-infra/cloud-workload-protection | cloud-workload-protection_tool.py | CWPP, container runtime security, serverless permission sprawl, escape detection |
Python Tools Reference
| Tool | Location | Key Flags | Output |
cloud-security-posture_tool.py | cloud-security-posture/scripts/ | --provider aws|azure|gcp, --account-id, --output json|table | CSPM findings with CIS check IDs, severity, compliance mapping, drift flag |
iac-security_tool.py | iac-security/scripts/ | --target terraform|cloudformation|kubernetes|helm, --dir, --output json | Misconfiguration findings with block_pr flag, compliance score |
container-image-scan_tool.py | container-image-scan/scripts/ | --image, --scanner trivy|grype|snyk, --output json|table | 11-field contract payload with per-finding component classification (base-image OS package / application dependency / unexpected layer) and block-deploy/fix-by-sla-window/track/accept action |
endpoint-os-security_tool.py | endpoint-os-security/scripts/ | --os windows|linux|macos, --host, --output json | Hardening score, EDR coverage status, CIS drift, patch priority matrix |
ot-iot-device-security_tool.py | ot-iot-device-security/scripts/ | --zone 0-5, --env ot|iot|mixed, --output json | Purdue zone risk assessment, segmentation gaps, compensating controls |
cloud-workload-protection_tool.py | cloud-workload-protection/scripts/ | --target containers|serverless, --cluster, --output json | CWPP coverage gaps, runtime anomalies, escape indicators |
All tools accept --help for full usage and --output json for machine-readable output.
# Quick invocation pattern for any tool in this domain
python <skill>/scripts/<skill>_tool.py --help
python <skill>/scripts/<skill>_tool.py --output jsonCloud Provider Coverage Matrix
| Tool | AWS | Azure | GCP | On-Prem |
cloud-security-posture_tool.py | Full (20+ CIS checks) | Full (10+ CIS checks) | Full (10+ CIS checks) | Not applicable |
iac-security_tool.py | Full (Terraform AWS provider, CloudFormation) | Partial (Terraform AzureRM) | Partial (Terraform Google) | On-prem Kubernetes |
endpoint-os-security_tool.py | EC2 / ECS task (agent-based) | Azure VM / AKS node | GCE instance / GKE node | Physical and VM endpoints |
ot-iot-device-security_tool.py | Not applicable | Not applicable | Not applicable | Full (all Purdue zones) |
cloud-workload-protection_tool.py | EKS, Lambda, ECS Fargate | AKS, Azure Functions | GKE, Cloud Run | Self-hosted Kubernetes |
Coverage levels:
- Full: All provider-specific checks implemented with benchmark IDs
- Partial: Core checks implemented; provider-specific hardening extensions in progress
- Not applicable: Tool is out of scope for that environment type
IaC Security Scanning
Supported IaC Formats
| Format | Toolchain | Scanner Backend | Notes |
| Terraform (HCL) | All providers | Checkov, tfsec | Modules analyzed recursively |
| CloudFormation (JSON/YAML) | AWS | Checkov, cfn-nag | SAM templates included |
| Pulumi | AWS, Azure, GCP | Checkov (Pulumi mode) | Python and TypeScript supported |
| Kubernetes manifests | All | Trivy config, kube-bench | Includes Kustomize overlays |
| Helm charts | All | Trivy config | Values file substitution applied before scan |
Check Categories
| Category | Description | Severity Range | CI/CD Gate |
| Public exposure | Resources publicly accessible without authentication | Critical | Block PR |
| IAM wildcard | Action: * with Resource: * in any policy | Critical | Block PR |
| Encryption at rest | Missing encryption configuration on storage, database, secrets | High | Block PR |
| Secret in IaC | Hardcoded credentials, API keys, certificates in HCL/YAML | Critical | Block PR |
| Network segmentation | SSH/RDP open to 0.0.0.0/0, unrestricted security groups | Critical–High | Block PR |
| Kubernetes pod security | privileged: true, hostPID, allowPrivilegeEscalation | Critical–High | Block PR |
| Kubernetes RBAC | Service accounts bound to cluster-admin | High | Block PR |
| Audit logging | CloudTrail, Azure Monitor, GCP Audit Logs disabled | High | Warn |
| Backup and versioning | Versioning disabled on object storage | Medium | Warn |
| Least privilege | Overly permissive RBAC, inline IAM policies | Medium | Warn |
Automated Scanner Commands
# Checkov — Terraform and CloudFormation
checkov -d ./terraform --framework terraform --output json
checkov -d ./cloudformation --framework cloudformation --output json
# Trivy — Kubernetes manifests and Helm charts
trivy config ./kubernetes/ --format json --severity HIGH,CRITICAL
trivy config ./helm-chart/ --format json --severity HIGH,CRITICAL
# tfsec — Terraform focused
tfsec . --format json
# kube-bench — CIS Kubernetes Benchmark (run on cluster node)
kube-bench run --targets node,master --jsonDomain Best Practices
- Never scope a posture scan to a single region or account. Cloud footprints span multiple regions, accounts, subscriptions, and projects. Cross-account S3 replication targets, VPC peering partners, and shared service accounts are all in scope. Single-region scans produce false confidence.
- Use configuration data no older than 24 hours for active posture assessments. Stale scan data masks drift. Drift events (unauthorized configuration changes) are themselves a High finding and must be reported separately from the configuration state.
- Apply severity modifiers consistently. Base severity from CIS Benchmarks is a starting point, not a final verdict. Internet-facing resources, production environment, and regulated data stores each warrant a severity upgrade. Verified compensating controls are the only valid basis for a severity downgrade.
- Shift IaC security left — block at PR, not at deployment. The
iac-securityskill is most effective as a CI/CD gate. A Critical IaC finding that reaches production is five to ten times more expensive to remediate than one caught in a PR comment. - Treat unauthorized drift as an independent finding. A configuration change that has no associated change management ticket is a process violation regardless of whether the resulting configuration is secure. Log it, alert on it, and require a retrospective ticket.
- Maintain separate OT and IT security procedures. Standard IT patching and hardening timelines do not apply to OT environments. In ICS/SCADA zones, availability and safety take priority over confidentiality. All recommendations for OT environments must include a compensating control path for cases where patching is not operationally feasible.
- Require co-approval for all OT environment changes. Any mutating action in an OT environment requires three approvals: security engineer (originator), CISO, and operations director. Safety-critical systems additionally require a safety review before any change window opens. This is enforced in the
ot-iot-device-securityruntime contract. - Measure CWPP coverage as a percentage. A CWPP coverage gap — workloads running without a runtime protection agent — is a High finding when below 80% and a Critical finding when below 50% in production. Coverage percentage should be tracked as a domain KPI and reported in
security-posture-score. - Link cloud posture findings to IaC policies. When the
cloud-security-postureskill detects a misconfiguration that originates from a Terraform or CloudFormation resource, it cascades to theiac-securityskill for policy-as-code rule creation. This prevents the same misconfiguration from reappearing after the next deployment. - Document remediation commands; never auto-execute them. All CLI remediation commands (AWS CLI, Azure CLI,
gsutil,kubectl) are included in skill output as reference for human engineers. The skills in this domain are read-only analyzers — no skill autonomously modifies cloud resources, OS configurations, or OT systems.
Workflow: Cloud Posture Assessment
This cross-skill workflow represents a complete infrastructure security assessment cycle. Each step may run independently or as part of the full sequence.
iac-security → cloud-security-posture → cloud-workload-protection → endpoint-os-securityStep-by-Step
Step 1 — IaC Security Gate (iac-security)
Trigger: PR opened or IaC file changed in CI/CD pipeline.
Action: Scan Terraform, CloudFormation, Kubernetes manifests, and Helm charts for misconfigurations. Block the PR if any Critical finding (public S3, IAM wildcard, secret in code, SSH open to 0.0.0.0/0) is present. Cascade passing results with compliance score to findings-tracker.
python cloud-infra/iac-security/scripts/iac-security_tool.py \
--target terraform --dir ./infrastructure --output jsonStep 2 — Cloud Posture Scan (cloud-security-posture) Trigger: Deployment event, daily scheduled scan, or IaC drift detected. Action: Collect current cloud resource configurations for all accounts, subscriptions, and projects. Run all CIS Benchmark checks. Apply severity modifiers. Detect drift against last baseline. Map findings to CIS, NIST 800-53, SOC 2, PCI DSS, HIPAA. Cascade Critical production findings to the USAP orchestrator immediately.
python cloud-infra/cloud-security-posture/scripts/cloud-security-posture_tool.py \
--provider aws --account-id 123456789012 --output jsonStep 3 — Workload Runtime Assessment (cloud-workload-protection)
Trigger: Completion of Step 2 or new container workload deployed.
Action: Assess CWPP tool coverage across all pods and serverless functions. Detect runtime anomalies and container escape indicators. Evaluate Lambda/Function permission sprawl. Escalate container escape events to incident-commander (SEV1) immediately.
python cloud-infra/cloud-workload-protection/scripts/cloud-workload-protection_tool.py \
--target containers --cluster prod-eks --output jsonStep 4 — Endpoint Hardening Check (endpoint-os-security) Trigger: Completion of Step 3 or new endpoint enrolled. Action: Evaluate hardening baseline compliance (CIS Level 1/2, DISA STIG). Assess EDR coverage and alert on gaps. Identify missing patches by CVSS score and CISA KEV status. Produce patch priority matrix with SLA windows.
python cloud-infra/endpoint-os-security/scripts/endpoint-os-security_tool.py \
--os linux --host prod-bastion-01 --output jsonStep 5 — OT Environment Assessment (ot-iot-device-security) (if in scope) Trigger: Scheduled quarterly OT assessment or OT incident. Action: Map all Purdue model zones. Identify segmentation gaps in the IT/OT DMZ. Assess default credentials on IoT devices. Flag safety-system risks as Critical regardless of other severity modifiers. All recommended actions include compensating controls and require co-approval.
python cloud-infra/ot-iot-device-security/scripts/ot-iot-device-security_tool.py \
--zone 3 --env ot --output jsonStep 6 — Aggregate and Report
Downstream: findings-tracker aggregates all findings. compliance-mapping generates framework-specific evidence packages. security-posture-score rolls up the domain posture score for executive reporting via ciso-brief-generator.
MITRE ATT&CK Cloud Matrix Coverage
The skills in this domain map to the MITRE ATT&CK Enterprise and ICS matrices as follows.
Initial Access
| Technique | ID | Detecting Skill |
| Valid Accounts — Cloud | T1078.004 | cloud-security-posture (IAM anomaly → cascade to identity-access-risk) |
| Exploit Public-Facing Application | T1190 | cloud-security-posture (internet-facing resources), cloud-workload-protection, container-image-scan (vulnerable base-image/dependency component reachable from a public-facing image) |
| Drive-by Compromise (HMI) | T1817 (ICS) | ot-iot-device-security |
| Default Credentials (IoT) | T1812 (ICS) | ot-iot-device-security |
Execution
| Technique | ID | Detecting Skill |
| Command and Scripting Interpreter — PowerShell | T1059.001 | endpoint-os-security (PS Script Block logging) |
| Container Administration Command | T1609 | cloud-workload-protection |
| Deploy Container | T1610 | cloud-workload-protection, iac-security |
| Modify Control Logic | T0833 (ICS) | ot-iot-device-security |
Persistence
| Technique | ID | Detecting Skill |
| Valid Accounts — Cloud | T1078.004 | cloud-security-posture, cloud-workload-protection |
| Implant Container Image | T1525 | cloud-workload-protection (image scanning), container-image-scan (unexpected/implanted layer detection against Dockerfile, CI build log, and SBOM) |
| Registry Run Keys / Startup Folder | T1547.001 | endpoint-os-security |
| Scheduled Task/Job | T1053.005 | endpoint-os-security |
Privilege Escalation
| Technique | ID | Detecting Skill |
| Abuse Elevation Control Mechanism | T1548 | endpoint-os-security |
| Container Escape | T1611 | cloud-workload-protection |
| Escape to Host | T1611 | cloud-workload-protection (privileged container detection) |
| Exploitation for Privilege Escalation | T1068 | endpoint-os-security (patch priority matrix) |
Defense Evasion
| Technique | ID | Detecting Skill |
| Modify Cloud Compute Infrastructure | T1578 | cloud-security-posture (drift detection) |
| Impair Defenses — Disable Cloud Logs | T1562.008 | cloud-security-posture (CloudTrail / Azure Monitor checks) |
| Impair Defenses — Disable or Modify Tools | T1562.001 | endpoint-os-security (EDR agent offline detection) |
Credential Access
| Technique | ID | Detecting Skill |
| OS Credential Dumping — LSASS | T1003.001 | endpoint-os-security (Credential Guard check) |
| Steal Application Access Token | T1528 | cloud-security-posture (IAM key rotation checks) |
| Forge Web Credentials | T1606 | cloud-security-posture (IAM policy analysis) |
Discovery
| Technique | ID | Detecting Skill |
| Cloud Service Discovery | T1526 | cloud-security-posture |
| Cloud Infrastructure Discovery | T1580 | cloud-security-posture |
| Container and Resource Discovery | T1613 | cloud-workload-protection |
| Network Service Discovery (OT) | T0840 (ICS) | ot-iot-device-security |
Lateral Movement
| Technique | ID | Detecting Skill |
| Use Alternate Authentication Material | T1550 | cloud-security-posture, endpoint-os-security |
| Lateral Tool Transfer | T1570 | endpoint-os-security |
| Remote Services — SSH | T1021.004 | endpoint-os-security, iac-security (SSH open check) |
| Lateral Movement — OT Network | T0812 (ICS) | ot-iot-device-security |
Impact
| Technique | ID | Detecting Skill |
| Data Destruction | T1485 | cloud-security-posture (S3 versioning, backup checks) |
| Defacement | T1491 | cloud-workload-protection |
| Endpoint Denial of Service | T1499 | endpoint-os-security |
| Modify Safety System | T0838 (ICS) | ot-iot-device-security (safety system impact flag) |
| Loss of Safety | T0879 (ICS) | ot-iot-device-security |
Compliance Framework Coverage
| Framework | Skills with Explicit Mapping |
| CIS AWS Foundations Benchmark | cloud-security-posture, iac-security |
| CIS Azure Benchmark | cloud-security-posture, iac-security |
| CIS Google Cloud Platform Benchmark | cloud-security-posture, iac-security |
| CIS Kubernetes Benchmark | iac-security, cloud-workload-protection |
| CIS Windows / Linux / macOS Benchmarks | endpoint-os-security |
| NIST SP 800-53 | cloud-security-posture, iac-security, endpoint-os-security |
| NIST SP 800-82 | ot-iot-device-security |
| NIST SP 800-213 (IoT) | ot-iot-device-security |
| IEC 62443 | ot-iot-device-security |
| NERC CIP | ot-iot-device-security |
| DISA STIG | endpoint-os-security |
| PCI DSS | cloud-security-posture, iac-security, endpoint-os-security |
| HIPAA | cloud-security-posture |
| SOC 2 | cloud-security-posture, iac-security |
Cascade and Integration Map
┌─────────────────────┐
│ iac-security │ ◄── CI/CD pipeline trigger
└────────┬────────────┘
│ IaC drift detected
▼
┌─────────────────────┐
│ cloud-security- │ ◄── daily scan / deployment event
│ posture │
└─────┬──────┬────────┘
│ │ public resource
│ └──► attack-surface-management
│ workload
▼
┌─────────────────────┐
│ cloud-workload- │
│ protection │ ──► incident-commander (escape)
└─────┬───────────────┘
│ endpoint in scope
▼
┌─────────────────────┐
│ endpoint-os- │
│ security │ ──► detection-engineering
└─────────────────────┘
┌─────────────────────┐
│ ot-iot-device- │ ◄── scheduled OT assessment
│ security │ ──► incident-commander (safety)
└─────────────────────┘
All skills ──► findings-tracker, compliance-mapping, security-posture-scoreRelated Domains
- appsec-devsecops/: Supplies IaC scan triggers from CI/CD pipeline events (
devsecops-pipeline,pipeline-security-scan). Theiac-securityskill consumes pipeline context and returns findings back to the AppSec domain'ssast-dast-coordinatorfor deduplication. - identity-access/: The
cloud-security-postureskill cascades IAM anomalies and wildcard policy findings toidentity-access-risk. Theendpoint-os-securityskill cascades local admin proliferation and credential guard gap findings toidentity-access-riskandcryptography-key-management.
Authoring Notes
When adding a new skill to this domain:
- Place the skill directory under
cloud-infra/<slug>/. - Reference the domain's Python tool naming convention:
<slug>_tool.pyinscripts/. - Update this CLAUDE.md Skills Catalog table and Cloud Provider Coverage Matrix.
- Update the root
README.mdDomain Index table entry forCloud & Infra. - Update
domains/cloud-infra.mdwith the new skill slug and level. - Ensure all MITRE ATT&CK technique mappings relevant to the new skill are added to the coverage section above.
- Add cascade rules to the Cascade and Integration Map.
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
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.

