| metadata | {"api_version":"azuresre.ai/v2","kind":"Skill"} |
| name | rbac-audit |
| description | Use when an operator or a scheduled task requests a periodic RBAC / least-privilege audit of the Zava Learning resource group — enumerate the real role assignments effective on the group, correlate each identity against actual Activity-Log usage to find standing access that is never used, flag over-privileged and directly-assigned (versus group-governed) identities and stale troubleshooting access, and recommend least-privilege corrections. Read-only; produces findings for a report. |
| tools | ["RunAzCliReadCommands","GetAzCliHelp","SearchMemory","ExecutePythonCode","microsoft-learn_microsoft_docs_search"] |
Zava Learning — RBAC / Least-Privilege & Access-Usage Audit
Resource Group: @@RG@@. Read-only audit — never create, modify, or remove a role assignment.
This is a weekly governance review of who can do what here, and whether they actually use it. The
goal is real least-privilege hygiene: surface over-privileged identities, direct (ungoverned)
assignments, and standing access that has gone unused so a human can clean it up.
Identity naming policy (governance reports name their subjects)
The principals under audit (their display name / UPN / objectId and role) are the subject of this
report — show them; an RBAC finding that hides the identity is useless. This is NOT the learner-PII
the redaction standard protects. Still apply SearchMemory("zava-redaction") redact() to strip any
secrets, tokens, credentials, or learner/customer PII that happen to appear, and never print
secret values — but DO name the admins, service principals, and groups being audited.
Where the names come from (do NOT call Microsoft Graph): take the identity name from the
principalName field that az role assignment list already returns, plus principalType. Do not
run az ad user/group/sp list|show or az rest https://graph.microsoft.com/... to resolve names or
look up guest/enabled status — the SRE Agent's managed identity has no Graph directory-read
permission, so those calls fail with Insufficient privileges AND are gated as PendingAuthorization,
which would stall this unattended weekly run. If a principalName is blank, identify that principal by
its principalId (objectId) + principalType and note "display name unavailable (directory lookup not
permitted)". A UPN containing #EXT# (when present in principalName) still indicates a guest.
1. Enumerate the real assignments effective on the RG
- All assignments scoped at or above the RG (this includes subscription/management-group inheritance —
most access here is inherited):
az role assignment list --scope /subscriptions/<sub>/resourceGroups/@@RG@@ --include-inherited -o json.
- For each: capture
principalName, principalType (User / Group / ServicePrincipal), roleDefinitionName,
scope, description, and createdOn where present. These fields are sufficient — everything you
need to name and classify a principal is already in this output; no follow-up directory calls.
2. Correlate against ACTUAL usage (the key step, BEST-EFFORT)
Standing access that is never exercised is the prime cleanup target. Pull who has actually operated in
this RG and join it to the assignment list:
- The
-g server-side filter on the activity log is unreliable here — query subscription-wide and
filter client-side on resourceGroupName == @@RG@@. Use a bounded, valid query (the API rejects
oversized requests with HTTP 400 "Bad Request"): a window of 30 days and --max-events 1000,
e.g. az monitor activity-log list --start-time <UTC now-30d> --end-time <UTC now> --max-events 1000 -o json.
If that still 400s, halve the window (e.g. 7-14 days) and/or add --query "[].{c:caller,r:resourceGroupName,t:eventTimestamp}".
Then keep events where resourceGroupName is the RG; aggregate distinct caller values (these are
UPNs or principal objectIds) with their last-seen timestamp.
- A principal that holds a privileged role but appears as no caller in the window is
standing-but-unused access → recommend removal / down-scoping. (Activity Log is control-plane,
~90-day retention; state the window you actually used and note data-plane-only usage may not appear.)
- Usage correlation is best-effort and MUST NOT block the report. If the activity-log query keeps
failing or returns nothing, do not retry indefinitely — after at most two adjusted attempts, record
"usage data unavailable (control-plane query limit / retention)" against the affected identities and
continue to the findings table and the branded deck using the assignment, scope, identity-type,
and
createdOn/description signals alone. Finishing the audit with a clear caveat is required;
stalling on usage correlation is a failure.
- Do not attempt Entra sign-in logs or any other Microsoft Graph query (e.g.
az rest https://graph.microsoft.com/...auditLogs/signIns) — the agent identity is not permitted and
the call would stall on approval / fail. Base usage purely on the Activity Log window above.
3. Flag findings with impact-based severity (see zava-audit-report model)
- SEV1 — Owner / User Access Administrator / Contributor held directly by a User (not via a
group), especially with no usage in the lookback window; any privileged role on a guest
(
#EXT# in principalName) or orphaned principal; a clearly temporary / break-glass grant still present (telltale
description like "temp", "break-glass", "INC-####", "troubleshooting", or an old createdOn).
- SEV2 — an identity assigned broader scope than needed (privileged role at subscription/MG
scope when only this RG is used), or a privileged role with stale usage (used long ago, not
recently); a direct privileged grant that should be group-governed and PIM-eligible.
- SEV3 — non-privileged direct assignment that should be group-governed; duplicate / redundant
assignments; a standing role that should be converted to a PIM eligible (just-in-time) assignment.
4. Recognise legitimate platform identities (avoid false positives)
This subscription carries first-party security & governance service principals — do not flag them as
risky; acknowledge them as expected. These include Microsoft Defender for Cloud, Azure Policy /
remediation, AzSecPack, Wiz (Wiz*), Illumio, Connected Machine / Arc, Backup / Recovery Services,
and the SRE Agent's own managed identity and the platform's user-assigned identity (AcrPull, Key Vault
Secrets User, Network Contributor for the agent). Note them in an "expected platform identities"
appendix rather than the findings table. Focus the findings on human users and non-platform
principals.
5. Produce the findings + recommendations
One row per finding:
Identity (type) · Role · Scope (direct/inherited) · Governance (direct/group) · Last used · Finding · Severity · Recommendation.
Sort SEV1 → SEV3. Recommendations should be concrete least-privilege moves: remove unused standing
access; replace a direct user grant with membership in a governed group; convert standing
privileged grants to PIM eligible (JIT); down-scope from subscription to the specific RG/resource;
remove stale break-glass access. Also produce a posture summary (Healthy / Needs attention / At risk,
counts by severity, # of identities with privileged-but-unused access, the single most important
action). Cite Azure least-privilege / PIM guidance via microsoft-learn_microsoft_docs_search.
6. Hand off to the branded report
This skill does not remediate. Pass the posture summary, the findings rows (each with its severity),
and the recommendations to the zava-audit-report skill, which renders the single branded,
downloadable PowerPoint deck and returns its download link. Surface that link to the operator.
