@raishin/vanguard-frontier-agentic 2.10.0 → 2.11.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@raishin/vanguard-frontier-agentic",
-  "version": "2.10.0",
+  "version": "2.11.0",
   "description": "Cloud and zero-trust agentic workflow marketplace for skills, agents, rules, MCP references, and compliance-aware architecture.",
   "license": "Apache-2.0",
   "repository": {
@@ -89,7 +89,7 @@
     "@semantic-release/release-notes-generator": "14.1.1",
     "conventional-changelog-conventionalcommits": "9.3.1",
     "fast-check": "^4.7.0",
-    "semantic-release": "25.0.3"
+    "semantic-release": "25.0.5"
   },
   "publishConfig": {
     "access": "public",

package/plugins/vanguard-frontier-agentic/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "vanguard-frontier-agentic",
-  "version": "2.10.0",
+  "version": "2.11.0",
   "description": "Curated marketplace for cloud and zero-trust AI workflows. 331 agents, 286 skills, and rules across AWS, Azure, OCI, GCP, Alibaba Cloud, Huawei Cloud, Kubernetes, and Terraform.",
   "author": {
     "name": "Raishin",

package/powers/README.md

@@ -1,6 +1,6 @@
 # `powers/` — Kiro Powers
 
-This directory holds **35 Kiro Powers** for `vanguard-frontier-agentic`, one
+This directory holds **38 Kiro Powers** for `vanguard-frontier-agentic`, one
 per cloud/platform/IaC provider. Each Power is a directory containing a
 `POWER.md` file with strict-5 frontmatter and steering content.
 
@@ -17,6 +17,7 @@ powers/
 ├── vanguard-cert-manager/POWER.md
 ├── vanguard-cilium/POWER.md
 ├── vanguard-contabo/POWER.md
+├── vanguard-databricks/POWER.md
 ├── vanguard-dotnet/POWER.md
 ├── vanguard-falco/POWER.md
 ├── vanguard-finance/POWER.md
@@ -32,6 +33,7 @@ powers/
 ├── vanguard-kyverno/POWER.md
 ├── vanguard-legal/POWER.md
 ├── vanguard-marketing/POWER.md
+├── vanguard-microsoft/POWER.md
 ├── vanguard-multi-cloud/POWER.md
 ├── vanguard-netsuite/POWER.md
 ├── vanguard-nvidia/POWER.md
@@ -42,6 +44,7 @@ powers/
 ├── vanguard-salesforce/POWER.md
 ├── vanguard-scaleway/POWER.md
 ├── vanguard-sigstore/POWER.md
+├── vanguard-snowflake/POWER.md
 └── vanguard-terraform/POWER.md
 ```
 
@@ -79,7 +82,7 @@ cd vanguard-frontier-agentic
 ## How to update
 
 ```bash
-# Regenerate the 35 Powers from catalog/agents.json + per-provider config:
+# Regenerate the 38 Powers from catalog/agents.json + per-provider config:
 npm run kiro-powers:write
 
 # Then verify everything is in sync:

package/powers/vanguard-databricks/POWER.md

@@ -0,0 +1,42 @@
+---
+name: "vanguard-databricks"
+displayName: "Vanguard Frontier — Databricks (Azure)"
+description: "Curated Azure Databricks agents for Unity Catalog governance and lakehouse engineering — static review only, no workspace or production mutations. Covers Unity Catalog securables and schema-scoped least-privilege grants, identity federation and account groups, account/workspace/metastore admin separation, run-production-as-service-principal, Entra ID managed identities, Access Connector + ADLS Gen2 external locations, cluster policies, and audit logs. Databricks and Azure surfaces are drift-prone; agents always verify against current Databricks and Microsoft Learn documentation before rendering findings."
+keywords: ["databricks", "azure", "unity-catalog", "lakehouse", "least-privilege", "data-engineering", "static-review"]
+author: "Raishin"
+---
+# Vanguard Frontier — Databricks (Azure)
+
+Curated Azure Databricks agents for Unity Catalog governance and lakehouse engineering — static review only, no workspace or production mutations. Covers Unity Catalog securables and schema-scoped least-privilege grants, identity federation and account groups, account/workspace/metastore admin separation, run-production-as-service-principal, Entra ID managed identities, Access Connector + ADLS Gen2 external locations, cluster policies, and audit logs. Databricks and Azure surfaces are drift-prone; agents always verify against current Databricks and Microsoft Learn documentation before rendering findings.
+
+## When to engage this Power
+
+Activate when the task references Databricks (Azure) services, resources, or operations. Do not activate on unrelated requests — narrow keyword matching is required to avoid false activations (Kiro Powers convention).
+
+## Routing pattern
+
+- *(no maestro for this provider; reference agents directly under `agents/databricks/`)*
+
+Reference agents directly from agents/databricks/ without maestro-based routing.
+
+## Live-guard agents (gate_mode only)
+
+- `databricks-live-unity-catalog-grant-guard-at-azure-agent` — never auto-dispatched; gate_mode only
+
+Live-guard agents enforce approval, target confirmation, evidence capture, and rollback plans before executing a mutation. They are never auto-dispatched — the maestro must place them in `live-guard-gate` or `runtime-evidence-gate` mode.
+
+## Invariants
+
+- Static review only — agents never request workspace tokens, service-principal secrets, storage keys, or customer data, and never mutate a Databricks workspace, Unity Catalog, or Azure resource.
+- Enforce least privilege: schema-scoped grants (CREATE TABLE/VOLUME/FUNCTION at schema level), no broad ALL PRIVILEGES, assign access to account groups not individuals, separate account/workspace/metastore admin roles.
+- Prefer Azure managed identities over service principals for storage access; production data is operated by service principals, not interactive users.
+- Production grant/role/policy/cluster changes are live-guard gated — never auto-dispatched; require explicit approval, scope confirmation, and rollback plan.
+
+## Where the agents live
+
+Agent specs and adapters are part of the [Vanguard Frontier Agentic](https://github.com/Raishin/vanguard-frontier-agentic) marketplace. For this provider, see `agents/databricks/` in that repository. All 3 agents in this provider ship a Kiro adapter (`harnesses/kiro-ide.agent.md`, `kiro-cli.agent.json`).
+
+## Companion install paths
+
+- **Claude Code:** `/plugin marketplace add Raishin/vanguard-frontier-agentic` then `/plugin install vanguard-frontier-agentic@vanguard-frontier-agentic`
+- **Codex / Copilot / Cursor / Gemini CLI / Kiro (file export):** `npx vfa-export-agents --platform <harness> --provider databricks --repo .`

package/powers/vanguard-microsoft/POWER.md

@@ -0,0 +1,45 @@
+---
+name: "vanguard-microsoft"
+displayName: "Vanguard Frontier — Microsoft"
+description: "Curated Microsoft 365 and Dynamics 365 agents for tenant governance, Entra identity and Conditional Access, Intune endpoints, Purview data security and compliance, Defender XDR, Teams/SharePoint/Exchange collaboration, Microsoft 365 Copilot readiness, Power Platform governance, and Dynamics 365 ERP/CRM (Finance, Supply Chain, Business Central, Sales, Customer Service, Field Service) — static review only, no tenant or production mutations. Routes via microsoft-maestro to M365, D365, Power Platform, and Copilot specialist agents. Microsoft licensing, certification, and API surfaces are drift-prone; agents always verify against current Microsoft Learn documentation before rendering findings."
+keywords: ["microsoft", "m365", "d365", "entra", "purview", "copilot", "power-platform", "static-review"]
+author: "Raishin"
+---
+# Vanguard Frontier — Microsoft
+
+Curated Microsoft 365 and Dynamics 365 agents for tenant governance, Entra identity and Conditional Access, Intune endpoints, Purview data security and compliance, Defender XDR, Teams/SharePoint/Exchange collaboration, Microsoft 365 Copilot readiness, Power Platform governance, and Dynamics 365 ERP/CRM (Finance, Supply Chain, Business Central, Sales, Customer Service, Field Service) — static review only, no tenant or production mutations. Routes via microsoft-maestro to M365, D365, Power Platform, and Copilot specialist agents. Microsoft licensing, certification, and API surfaces are drift-prone; agents always verify against current Microsoft Learn documentation before rendering findings.
+
+## When to engage this Power
+
+Activate when the task references Microsoft services, resources, or operations. Do not activate on unrelated requests — narrow keyword matching is required to avoid false activations (Kiro Powers convention).
+
+## Routing pattern
+
+- **`copilot-governance-maestro-agent`** — classifies and routes the task to the right specialist
+
+Use the maestro as the entry point: classify the task, then dispatch to one specialist or a parallel team of specialists. Never have the maestro itself execute a live mutation.
+
+## Live-guard agents (gate_mode only)
+
+- `d365-live-record-field-update-guard-agent` — never auto-dispatched; gate_mode only
+- `d365-live-security-role-guard-agent` — never auto-dispatched; gate_mode only
+- `m365-live-identity-posture-guard-agent` — never auto-dispatched; gate_mode only
+- `m365-live-sensitivity-label-apply-guard-agent` — never auto-dispatched; gate_mode only
+
+Live-guard agents enforce approval, target confirmation, evidence capture, and rollback plans before executing a mutation. They are never auto-dispatched — the maestro must place them in `live-guard-gate` or `runtime-evidence-gate` mode.
+
+## Invariants
+
+- Static review only — agents never request tenant credentials, tokens, customer data, or PII, and never mutate a Microsoft 365 tenant or Dynamics 365 environment.
+- Apply Zero Trust by default: verify explicitly, least privilege (JIT/JEA), assume breach; confirm tenant, environment, and data classification before any recommendation.
+- Microsoft 365 Copilot and Copilot Studio configurations are adversarially reviewed for oversharing, ungrounded Graph exposure, and missing human-handoff controls before any approve decision.
+- Production-impacting actions (Conditional Access changes, D365 cutover, Power Platform prod deploy, MFA changes) are live-guard gated — never auto-dispatched; require explicit approval, scope confirmation, and rollback plan.
+
+## Where the agents live
+
+Agent specs and adapters are part of the [Vanguard Frontier Agentic](https://github.com/Raishin/vanguard-frontier-agentic) marketplace. For this provider, see `agents/microsoft/` in that repository. All 40 agents in this provider ship a Kiro adapter (`harnesses/kiro-ide.agent.md`, `kiro-cli.agent.json`).
+
+## Companion install paths
+
+- **Claude Code:** `/plugin marketplace add Raishin/vanguard-frontier-agentic` then `/plugin install vanguard-frontier-agentic@vanguard-frontier-agentic`
+- **Codex / Copilot / Cursor / Gemini CLI / Kiro (file export):** `npx vfa-export-agents --platform <harness> --provider microsoft --repo .`

package/powers/vanguard-snowflake/POWER.md

@@ -0,0 +1,42 @@
+---
+name: "vanguard-snowflake"
+displayName: "Vanguard Frontier — Snowflake (Azure)"
+description: "Curated Snowflake-on-Azure agents for RBAC access governance and data-platform engineering — static review only, no account or production mutations. Covers role hierarchy and least privilege, ACCOUNTADMIN restriction, SECURITYADMIN/SYSADMIN separation of duties, future grants and managed-access schemas, network policies, key-pair/Entra OAuth/SSO/SCIM authentication, Azure Private Link and storage integration, and masking/row-access/tagging governance. Snowflake and Azure surfaces are drift-prone; agents always verify against current Snowflake and Microsoft Learn documentation before rendering findings."
+keywords: ["snowflake", "azure", "rbac", "least-privilege", "data-governance", "private-link", "static-review"]
+author: "Raishin"
+---
+# Vanguard Frontier — Snowflake (Azure)
+
+Curated Snowflake-on-Azure agents for RBAC access governance and data-platform engineering — static review only, no account or production mutations. Covers role hierarchy and least privilege, ACCOUNTADMIN restriction, SECURITYADMIN/SYSADMIN separation of duties, future grants and managed-access schemas, network policies, key-pair/Entra OAuth/SSO/SCIM authentication, Azure Private Link and storage integration, and masking/row-access/tagging governance. Snowflake and Azure surfaces are drift-prone; agents always verify against current Snowflake and Microsoft Learn documentation before rendering findings.
+
+## When to engage this Power
+
+Activate when the task references Snowflake (Azure) services, resources, or operations. Do not activate on unrelated requests — narrow keyword matching is required to avoid false activations (Kiro Powers convention).
+
+## Routing pattern
+
+- *(no maestro for this provider; reference agents directly under `agents/snowflake/`)*
+
+Reference agents directly from agents/snowflake/ without maestro-based routing.
+
+## Live-guard agents (gate_mode only)
+
+- `snowflake-live-rbac-grant-guard-at-azure-agent` — never auto-dispatched; gate_mode only
+
+Live-guard agents enforce approval, target confirmation, evidence capture, and rollback plans before executing a mutation. They are never auto-dispatched — the maestro must place them in `live-guard-gate` or `runtime-evidence-gate` mode.
+
+## Invariants
+
+- Static review only — agents never request account credentials, key-pair private keys, OAuth secrets, or customer data, and never mutate a Snowflake account or Azure resource.
+- Enforce least privilege: custom business-function roles under SYSADMIN, restrict ACCOUNTADMIN to a minimum of controlled users, never grant sensitive privileges to PUBLIC, separate SECURITYADMIN (grants) from SYSADMIN (objects).
+- Service accounts use key-pair or Entra OAuth (never passwords); enforce network policies and MFA for human users.
+- Production role/grant/policy/warehouse changes are live-guard gated — never auto-dispatched; require explicit approval, scope confirmation, and rollback plan.
+
+## Where the agents live
+
+Agent specs and adapters are part of the [Vanguard Frontier Agentic](https://github.com/Raishin/vanguard-frontier-agentic) marketplace. For this provider, see `agents/snowflake/` in that repository. All 3 agents in this provider ship a Kiro adapter (`harnesses/kiro-ide.agent.md`, `kiro-cli.agent.json`).
+
+## Companion install paths
+
+- **Claude Code:** `/plugin marketplace add Raishin/vanguard-frontier-agentic` then `/plugin install vanguard-frontier-agentic@vanguard-frontier-agentic`
+- **Codex / Copilot / Cursor / Gemini CLI / Kiro (file export):** `npx vfa-export-agents --platform <harness> --provider snowflake --repo .`

package/schemas/agent.schema.json

@@ -56,7 +56,10 @@
         "salesforce",
         "netsuite",
         "accounting",
-        "finance"
+        "finance",
+        "microsoft",
+        "databricks",
+        "snowflake"
       ]
     },
     "harnesses": {

package/schemas/mcp-reference.schema.json

@@ -42,7 +42,10 @@
         "kubernetes",
         "terraform",
         "multi-cloud",
-        "generic"
+        "generic",
+        "microsoft",
+        "databricks",
+        "snowflake"
       ]
     },
     "harnesses": {

package/schemas/rule.schema.json

@@ -37,7 +37,10 @@
         "kubernetes",
         "terraform",
         "multi-cloud",
-        "generic"
+        "generic",
+        "microsoft",
+        "databricks",
+        "snowflake"
       ]
     },
     "harnesses": {

package/schemas/skill.schema.json

@@ -47,7 +47,10 @@
         "hr",
         "legal",
         "salesforce",
-        "netsuite"
+        "netsuite",
+        "microsoft",
+        "databricks",
+        "snowflake"
       ]
     },
     "harnesses": {

package/scripts/generate-docs-data.mjs

@@ -58,6 +58,8 @@ const taxonomy = [
   { category: 'Developer Platforms', providers: ['backstage', 'dotnet', 'generic', 'multi-cloud'] },
   { category: 'ERP & Finance', providers: ['netsuite', 'accounting', 'finance'] },
   { category: 'Business Functions', providers: ['salesforce', 'legal', 'hr', 'marketing'] },
+  { category: 'Microsoft 365 & Dynamics 365', providers: ['microsoft'] },
+  { category: 'Data & Analytics Platforms', providers: ['databricks', 'snowflake'] },
 ];
 
 const taxonomyYaml = taxonomy.map(t => {

package/scripts/generate-kiro-powers.mjs

@@ -208,6 +208,42 @@ const PROVIDERS = {
       "Live-guard agent (salesforce-live-guard-agent) must never be auto-dispatched; require explicit approval, target org confirmation, and rollback plan.",
     ],
   },
+  microsoft: {
+    displayName: "Vanguard Frontier — Microsoft",
+    description:
+      "Curated Microsoft 365 and Dynamics 365 agents for tenant governance, Entra identity and Conditional Access, Intune endpoints, Purview data security and compliance, Defender XDR, Teams/SharePoint/Exchange collaboration, Microsoft 365 Copilot readiness, Power Platform governance, and Dynamics 365 ERP/CRM (Finance, Supply Chain, Business Central, Sales, Customer Service, Field Service) — static review only, no tenant or production mutations. Routes via microsoft-maestro to M365, D365, Power Platform, and Copilot specialist agents. Microsoft licensing, certification, and API surfaces are drift-prone; agents always verify against current Microsoft Learn documentation before rendering findings.",
+    keywords: ["microsoft", "m365", "d365", "entra", "purview", "copilot", "power-platform", "static-review"],
+    invariants: [
+      "Static review only — agents never request tenant credentials, tokens, customer data, or PII, and never mutate a Microsoft 365 tenant or Dynamics 365 environment.",
+      "Apply Zero Trust by default: verify explicitly, least privilege (JIT/JEA), assume breach; confirm tenant, environment, and data classification before any recommendation.",
+      "Microsoft 365 Copilot and Copilot Studio configurations are adversarially reviewed for oversharing, ungrounded Graph exposure, and missing human-handoff controls before any approve decision.",
+      "Production-impacting actions (Conditional Access changes, D365 cutover, Power Platform prod deploy, MFA changes) are live-guard gated — never auto-dispatched; require explicit approval, scope confirmation, and rollback plan.",
+    ],
+  },
+  databricks: {
+    displayName: "Vanguard Frontier — Databricks (Azure)",
+    description:
+      "Curated Azure Databricks agents for Unity Catalog governance and lakehouse engineering — static review only, no workspace or production mutations. Covers Unity Catalog securables and schema-scoped least-privilege grants, identity federation and account groups, account/workspace/metastore admin separation, run-production-as-service-principal, Entra ID managed identities, Access Connector + ADLS Gen2 external locations, cluster policies, and audit logs. Databricks and Azure surfaces are drift-prone; agents always verify against current Databricks and Microsoft Learn documentation before rendering findings.",
+    keywords: ["databricks", "azure", "unity-catalog", "lakehouse", "least-privilege", "data-engineering", "static-review"],
+    invariants: [
+      "Static review only — agents never request workspace tokens, service-principal secrets, storage keys, or customer data, and never mutate a Databricks workspace, Unity Catalog, or Azure resource.",
+      "Enforce least privilege: schema-scoped grants (CREATE TABLE/VOLUME/FUNCTION at schema level), no broad ALL PRIVILEGES, assign access to account groups not individuals, separate account/workspace/metastore admin roles.",
+      "Prefer Azure managed identities over service principals for storage access; production data is operated by service principals, not interactive users.",
+      "Production grant/role/policy/cluster changes are live-guard gated — never auto-dispatched; require explicit approval, scope confirmation, and rollback plan.",
+    ],
+  },
+  snowflake: {
+    displayName: "Vanguard Frontier — Snowflake (Azure)",
+    description:
+      "Curated Snowflake-on-Azure agents for RBAC access governance and data-platform engineering — static review only, no account or production mutations. Covers role hierarchy and least privilege, ACCOUNTADMIN restriction, SECURITYADMIN/SYSADMIN separation of duties, future grants and managed-access schemas, network policies, key-pair/Entra OAuth/SSO/SCIM authentication, Azure Private Link and storage integration, and masking/row-access/tagging governance. Snowflake and Azure surfaces are drift-prone; agents always verify against current Snowflake and Microsoft Learn documentation before rendering findings.",
+    keywords: ["snowflake", "azure", "rbac", "least-privilege", "data-governance", "private-link", "static-review"],
+    invariants: [
+      "Static review only — agents never request account credentials, key-pair private keys, OAuth secrets, or customer data, and never mutate a Snowflake account or Azure resource.",
+      "Enforce least privilege: custom business-function roles under SYSADMIN, restrict ACCOUNTADMIN to a minimum of controlled users, never grant sensitive privileges to PUBLIC, separate SECURITYADMIN (grants) from SYSADMIN (objects).",
+      "Service accounts use key-pair or Entra OAuth (never passwords); enforce network policies and MFA for human users.",
+      "Production role/grant/policy/warehouse changes are live-guard gated — never auto-dispatched; require explicit approval, scope confirmation, and rollback plan.",
+    ],
+  },
 };
 
 const catalog = JSON.parse(readFileSync(catalogPath, "utf8"));

package/scripts/release-prepare.mjs

@@ -60,6 +60,41 @@ const VERSION_PINNED_NESTED = [
   { path: ".claude-plugin/marketplace.json", key: "metadata.version" },
 ];
 
+// Write the resolved release version into package.json BEFORE any derived
+// artifact is generated.
+//
+// @semantic-release/npm's prepare step (which runs `npm version <v>
+// --no-git-tag-version --allow-same-version`) is ordered AFTER this exec
+// step, so at this point package.json still holds the PREVIOUS version.
+// The generators below (generate-plugin-manifest / generate-cursor-plugin)
+// and the asset-integrity hash all derive from package.json.version; without
+// this write they would read the old version and silently revert the
+// Claude/Cursor manifests — and stale the package.json hash — every release.
+//
+// Uses a minimal, format-preserving text edit (not JSON.parse/stringify) so
+// the result is byte-identical to what `npm version --allow-same-version`
+// produces afterwards. That keeps npm's later run a true no-op and prevents
+// it from re-staling the asset-integrity manifest generated here. The
+// top-level "version" key is the only `"version":` key in package.json
+// (dependency entries are keyed by package name), so the first match is the
+// package version.
+function syncPackageJsonVersion(nextVersion) {
+  const abs = join(REPO, "package.json");
+  const raw = readFileSync(abs, "utf8");
+  const updated = raw.replace(
+    /("version"\s*:\s*")[^"]+(")/,
+    `$1${nextVersion}$2`,
+  );
+  if (updated === raw) return false;
+  if (JSON.parse(updated).version !== nextVersion) {
+    throw new Error(
+      `[release-prepare] failed to write version ${nextVersion} into package.json`,
+    );
+  }
+  writeFileSync(abs, updated, "utf8");
+  return true;
+}
+
 function syncPluginVersion(relPath) {
   const abs = join(REPO, relPath);
   const data = JSON.parse(readFileSync(abs, "utf8"));
@@ -178,6 +213,11 @@ function regenerate(cmd, args) {
 
 console.log(`[release-prepare] syncing artifacts to version ${NEXT_VERSION}`);
 
+// Must happen first: every derived artifact below reads package.json.version.
+if (syncPackageJsonVersion(NEXT_VERSION)) {
+  console.log(`[release-prepare] updated package.json version -> ${NEXT_VERSION}`);
+}
+
 let touched = 0;
 for (const rel of VERSION_PINNED_PLUGINS) {
   if (syncPluginVersion(rel)) {

package/skills/cross-functional/audit-evidence-mapping-protocol/SKILL.md

@@ -0,0 +1,92 @@
+---
+name: audit-evidence-mapping-protocol
+description: Use this skill when compliance controls must be mapped to audit evidence, when evidence collection, retention policy, and legal-hold status need to be assessed across Microsoft 365 workloads, or when an attestation package must be assembled for an auditor or regulator. Defines the end-to-end flow from control identification through evidence discovery, gap analysis, retention verification, legal-hold confirmation, and attestation sign-off. Does not serve as an authorization to release evidence to external parties; that requires a human compliance or legal owner. Does not replace qualified legal or audit counsel.
+allowed-tools: Read Grep Glob
+metadata:
+  author: "github: Raishin"
+  version: "0.1.0"
+  updated: "2026-06-16"
+  category: compliance
+  lifecycle: experimental
+---
+
+# Audit Evidence Mapping Protocol
+
+## Purpose
+This skill defines how compliance controls are traced to concrete audit evidence, how that evidence is validated for completeness and retention, and how an attestation package is assembled and handed off to the compliance or legal owner. It exists so that evidence gaps are surfaced before an audit begins, retention and legal-hold status are verified against Microsoft Purview policy, and every handoff carries a documented chain of custody. It does not authorize evidence release, does not replace legal or audit counsel, and does not modify retention policies or legal-hold configuration.
+
+## When to use
+- A compliance or audit team must map controls to evidence artifacts across Microsoft 365 workloads.
+- Audit log retention status needs to be verified against organizational policy before a regulatory review.
+- A legal hold must be confirmed or flagged as incomplete before an eDiscovery matter proceeds.
+- An attestation package must be assembled from distributed evidence sources (Exchange, SharePoint, Entra ID, Dataverse).
+- Evidence completeness gaps are suspected and must be inventoried before the auditor window opens.
+
+## When NOT to use
+- Evidence has already been assembled and validated — hand off directly to the compliance owner.
+- The matter is an active legal proceeding requiring litigation-hold decisions — escalate to legal counsel immediately.
+- A retention policy change or legal-hold placement is required — that is a production-impacting action and must be escalated to the Purview compliance administrator and the legal owner.
+- The matter involves special-category personal data and jurisdiction is unknown — stop and escalate.
+
+## Participating agents
+- `d365-security-sod-governance-agent` — primary: maps segregation-of-duties controls to evidence artifacts in Dynamics 365 and Power Platform audit logs
+- `m365-maestro-agent` — orchestrates cross-workload evidence discovery; escalation path to Microsoft Purview specialists (planned: purview-compliance-specialist-agent)
+
+## Inputs required
+- Control framework or control list (e.g., ISO 27001, SOC 2, NIST CSF control IDs)
+- Workload scope (Exchange Online, SharePoint, Entra ID, Dynamics 365, Power Platform, Teams)
+- Retention requirements (regulatory, contractual, or internal policy)
+- Legal-hold case reference (if applicable)
+- Audit window dates
+
+## Evidence required
+- Existing audit log retention policies from Microsoft Purview (Audit Standard or Audit Premium tier)
+- Current legal-hold case list and hold scope from Microsoft Purview eDiscovery
+- Control register or risk register from the compliance team
+- Prior audit findings or evidence gaps (if available)
+
+## Workflow
+
+1. **Ingest control framework** — receive the control list and map each control to a workload and evidence type (audit log, configuration export, access review, policy document).
+2. **Scope workloads** — confirm which Microsoft 365 services and Dynamics 365 environments are in scope; note licensing tier (Audit Standard vs. Audit Premium) as it affects default retention (180 days vs. 1 year vs. 10 years with add-on).
+3. **Discover evidence artifacts** — for each control, identify the evidence artifact location (Purview Audit log, SharePoint document library, Entra ID access review, Power Platform DLP report).
+4. **Verify retention status** — confirm that audit log retention policies cover the required period; flag any control whose evidence falls under the 180-day standard default that requires a longer period.
+5. **Check legal-hold coverage** — if a legal hold is active, confirm that evidence in scope is covered by an eDiscovery hold; flag uncovered artifacts.
+6. **Identify evidence gaps** — produce a gap register: controls with missing artifacts, expired retention, or uncovered legal-hold scope.
+7. **Escalation gate: evidence completeness** — if gap register is non-empty, pause and escalate to compliance owner and m365-maestro-agent for Purview remediation.
+8. **Escalation gate: retention or legal-hold deficiency** — if retention is insufficient or a legal hold is not in place, stop and escalate to the Purview compliance administrator and legal owner before proceeding.
+9. **Assemble attestation package** — for each control, compile: control ID, evidence artifact reference, evidence location, retention expiry, legal-hold status, evidence quality rating (complete / partial / missing), and open questions.
+10. **Review for privilege and privacy sensitivity** — label package sections with privilege sensitivity and privacy sensitivity before handoff.
+11. **Hand off to compliance owner** — deliver attestation package with a do-not-do list and open questions; require human sign-off before any evidence is transmitted to an external auditor or regulator.
+
+## Decision gates
+
+| Gate | Condition | Action |
+|---|---|---|
+| Evidence completeness | Any control has missing or partial evidence | Pause; escalate to compliance owner; open remediation task |
+| Retention / legal-hold | Retention < required period OR legal hold not confirmed | Stop; escalate to Purview admin and legal owner before proceeding |
+| Privilege sensitivity | Evidence touches legally privileged communications | Flag; route to legal counsel before inclusion in package |
+| Special-category data | Evidence includes health, biometric, or regulated personal data | Stop; confirm jurisdiction and privacy owner before proceeding |
+
+## Refusal triggers
+- A request is made to delete, truncate, or shorten retention of in-scope evidence — refuse and escalate.
+- A request is made to release evidence directly to an external auditor without human compliance owner sign-off — refuse.
+- Credentials, tenant IDs, or personal data are requested to perform evidence discovery — refuse; work from sanitized signals only.
+- Jurisdiction of regulated personal data is unknown — refuse to proceed; escalate.
+
+## Handoff rules
+- Every handoff carries: control ID list, evidence gap register, retention status summary, legal-hold status, privilege sensitivity label, privacy sensitivity label, open questions, and a do-not-do list.
+- No agent transmits evidence to external parties. Human compliance owner authorizes all external transmissions.
+- Post-handoff, the receiving agent or human confirms receipt and records the handoff in the audit log.
+
+## KPIs
+- Percentage of controls with complete, verified evidence before audit window opens
+- Number of retention gaps identified and remediated before audit
+- Time from control mapping to attestation package delivery
+- Number of escalations requiring human intervention
+
+## References
+- [Learn about auditing solutions in Microsoft Purview](https://learn.microsoft.com/purview/audit-solutions-overview)
+- [Manage audit log retention policies](https://learn.microsoft.com/purview/audit-log-retention-policies)
+- [Microsoft Purview eDiscovery solutions](https://learn.microsoft.com/en-us/purview/ediscovery)
+- [Microsoft Purview Data Lifecycle Management](https://learn.microsoft.com/en-us/purview/data-lifecycle-management)

package/skills/cross-functional/audit-evidence-mapping-protocol/metadata.json

@@ -0,0 +1,20 @@
+{
+  "id": "audit-evidence-mapping-protocol",
+  "name": "Audit Evidence Mapping Protocol",
+  "type": "skill",
+  "provider": "generic",
+  "harnesses": ["codex", "claude-code", "cursor", "gemini", "kiro", "other"],
+  "summary": "Maps compliance controls to audit evidence artifacts across Microsoft 365 workloads, verifies retention and legal-hold status against Microsoft Purview policy, identifies evidence gaps before the audit window opens, and assembles a signed-off attestation package. Covers Audit Standard and Audit Premium retention tiers (180 days, 1 year, 10 years with add-on), eDiscovery legal-hold verification, and privilege and privacy sensitivity labeling of the evidence package.",
+  "source_type": "original",
+  "official_docs": [
+    "https://learn.microsoft.com/purview/audit-solutions-overview",
+    "https://learn.microsoft.com/purview/audit-log-retention-policies",
+    "https://learn.microsoft.com/en-us/purview/ediscovery",
+    "https://learn.microsoft.com/en-us/purview/data-lifecycle-management"
+  ],
+  "security_notes": "Protocol is recommendation and orchestration only — never an authorization to release evidence to external parties or to modify retention policies, legal holds, or eDiscovery cases. All production-impacting steps (retention policy changes, legal-hold placement, evidence transmission to external auditors) must be escalated to the Purview compliance administrator and the human legal or compliance owner. Never requests credentials, tenant IDs, session tokens, or customer personal data to perform evidence discovery; works from sanitized control and workload scope signals only. Evidence touching legally privileged communications is flagged and routed to legal counsel before inclusion in any package. Special-category personal data triggers a jurisdiction confirmation gate before proceeding.",
+  "last_verified": "2026-06-16",
+  "path": "skills/cross-functional/audit-evidence-mapping-protocol",
+  "author": "github: Raishin",
+  "version": "0.1.0"
+}