Orca Account Health

Orca Account Health Skill

Answers the question: "Is every account connected, recently synced, and fully scanned — so we can trust the data we're about to act on?"

Blind spots in account coverage produce blind spots in every other skill: if prod-eu-west last synced 48 hours ago, the alerts you triage are stale. This skill audits the foundation.

Usage

/orca-account-health
/orca-account-health aws
/orca-account-health degraded

Or natural language:

• "are all accounts connected?"
• "is everything synced?"
• "coverage audit before the audit"
• "any blind spots?"
• "show me account health"

Processing Logic

Step 1: Determine Scope

Parse the input:

• No arg: full inventory across all clouds.
• Provider arg (aws, gcp, azure, kubernetes, oci): filter to one cloud.
• Status arg (degraded, offline, healthy): filter by health.

Step 2: Pull integration configs (single call)

get_integration_configs_data:

This returns every configured integration / source: cloud accounts, log sources (CloudTrail, Azure Monitor, etc.), Kubernetes clusters, and their extraction settings. Parse out:

• Source type and provider
• Connection state
• Last successful run / sync timestamp
• Any error messages or degraded indicators

Step 3: Cross-check coverage with discovery (parallel)

In parallel with Step 2, run two discovery_search queries to spot-check that data is actually flowing — not just that the connection exists:

Query A: Asset freshness per account

discovery_search:
  search_phrase: "assets grouped by cloud account"
  limit: 50

Query B: Recent CDR event flow (confirms log ingestion, not just inventory)

discovery_search:
  search_phrase: "cloud audit events in the last 24 hours grouped by account"
  limit: 50

If an account appears in get_integration_configs_data but returns zero assets or zero recent CDR events, that's a covert blind spot even though the integration reports "connected."

Step 4: Compute per-account health

For each account, derive a health verdict:

Step 5: Detect specific blind-spot patterns

Flag these patterns explicitly — they're the ones that bite during audits or investigations:

1. Account is connected but produces no assets → onboarding likely incomplete.
2. Account is connected, has assets, but no CDR events in 24h → audit log integration broken (CloudTrail / Azure Monitor / GCP Audit Logs misrouted).
3. Scanner deployed but stuck on a stale scan → SideScanning or workload scanner needs restart.
4. New account added < 7 days ago, no successful sync yet → onboarding incomplete or failed silently.
5. Provider drift — account listed in one cloud's console but missing from Orca.

Proactive Behavior

This skill is diagnostic, not remediation-heavy — but it should still surface the next step:

1. Identify the worst blind spot and recommend a specific fix:
   • "Reconnect prod-eu-west — last sync 48h ago, no CDR events. Open Orca Console → Cloud Accounts → Reconnect."
2. For onboarding failures, link to the docs:
   • /orca-account-health should reference documentation_search for "onboarding troubleshooting" if user asks "how do I fix it".
3. Pre-audit framing: if the user said "before the audit", explicitly say whether they can proceed or need to wait for re-sync.

Output Format

Layer 1: Coverage Dashboard

═══════════════════════════════════════════════════════════════════
ACCOUNT HEALTH — Cloud Coverage Audit
<date> | <scope>
═══════════════════════════════════════════════════════════════════

VERDICT: <one-liner — "Full coverage, all green" / "1 offline, 2 sync lag, action needed">

┌─────────────────────────────────────────────────────────────────┐
│  TOTAL ACCOUNTS    <N>                                          │
│  HEALTHY           <N>                                          │
│  SYNC LAG          <N> (24-72h since last sync)                 │
│  STALE             <N> (> 72h since last sync)                  │
│  DEGRADED          <N> (errors / scanner gap / no CDR)          │
│  OFFLINE           <N> (broken / never synced)                  │
│  RECENTLY ADDED    <N> (onboarded < 7d, verify completion)      │
│  BLIND SPOTS       <N> (connected but no data flowing)          │
└─────────────────────────────────────────────────────────────────┘

ACCOUNT STATUS:
  Account              Provider   Last Sync     Scanner   CDR    Status
  ────────────────────────────────────────────────────────────────────────
  <account-1>          AWS        2 min ago     YES       OK     Healthy
  <account-2>          AWS        26 hours ago  YES       OK     Sync Lag
  <account-3>          Azure      Never         NO        —      Offline
  <account-4>          GCP        2h ago        YES       NO     Degraded
  ...

BLIND SPOTS (immediate attention):
  [!] <account> — <specific issue>
      Impact: <what's invisible because of this gap>
      Fix: <specific action>

  [!] <account> — ...

RECOMMENDED ACTION:
  Reconnect <account> first — <reason>.
  After reconnect, re-run any pending investigation or audit.

═══════════════════════════════════════════════════════════════════
Or drill down: degraded | offline | sync-lag | recently-added |
by-provider | full
═══════════════════════════════════════════════════════════════════

Layer 2: Drill-Downs

"degraded" — Accounts with active problems

───────────────────────────────────────────────────────────────────
DEGRADED ACCOUNTS — Active Issues
───────────────────────────────────────────────────────────────────

  [!] <account> (<provider>)
      Connection: <state>
      Last sync: <time>
      Scanner: <deployed / missing / stuck>
      CDR ingestion: <events in 24h or "none">
      Error: <reported error message>
      Impact: <what's invisible — e.g. "any investigation in this
              account is missing the last 6h of CDR data">
      Fix: <specific action — link to documentation_search if needed>

  [!] <account> — ...

───────────────────────────────────────────────────────────────────

"offline" — Broken or never-synced accounts

───────────────────────────────────────────────────────────────────
OFFLINE ACCOUNTS
───────────────────────────────────────────────────────────────────

These accounts are configured but Orca has no data:

  [X] <account> (<provider>)
      Last successful sync: <date or NEVER>
      Connection error: <message>
      Likely cause: <inferred — e.g. revoked role, deleted trust policy>
      Fix path: <action>

───────────────────────────────────────────────────────────────────

"sync-lag" — Accounts behind on sync

───────────────────────────────────────────────────────────────────
SYNC LAG — 24-72h Behind
───────────────────────────────────────────────────────────────────

  <account> (<provider>) — last sync <time> ago
    Why it matters: <e.g. "alerts triaged here may be missing the
                    last day of activity">

  ...

If urgency is high (audit, investigation), force a re-sync:
  Orca Console → Cloud Accounts → <account> → Sync Now
───────────────────────────────────────────────────────────────────

"recently-added" — New accounts (verify onboarding)

───────────────────────────────────────────────────────────────────
RECENTLY ADDED ACCOUNTS — Verify Onboarding
───────────────────────────────────────────────────────────────────

  <account> (<provider>) — added <N> days ago
    First sync: <success / pending / failed>
    Assets discovered: <N>
    Scanner deployed: <Y/N>
    CDR ingestion: <active / pending>
    Onboarding status: <complete / pending / failed>

───────────────────────────────────────────────────────────────────

"by-provider" — Coverage split by cloud

───────────────────────────────────────────────────────────────────
COVERAGE BY PROVIDER
───────────────────────────────────────────────────────────────────

  Provider     Connected   Healthy   Issues
  ───────────────────────────────────────────
  AWS          <N>         <N>       <N>
  Azure        <N>         <N>       <N>
  GCP          <N>         <N>       <N>
  Kubernetes   <N>         <N>       <N>
  OCI          <N>         <N>       <N>

───────────────────────────────────────────────────────────────────

"full"

Show all sections in order.

Edge Cases

No Accounts Configured

⚠ No cloud accounts configured in Orca.

To get started:
  Orca Console → Cloud Accounts → Add Account
  Choose: AWS / Azure / GCP / Kubernetes / OCI

After onboarding, re-run /orca-account-health to verify coverage.

All Accounts Healthy

✅ All <N> accounts healthy.

  • All connections active
  • All accounts synced within the last hour
  • All scanners deployed
  • CDR ingestion active across all accounts

You're clear to proceed with audit / investigation / review.

Coverage Mismatch (account in cloud console but not in Orca)

⚠ Provider-side accounts NOT in Orca:

This skill can only see what's configured in Orca. If your AWS Org has
30 accounts but Orca has 25, the other 5 are invisible here.

To verify completeness:
  • Compare Orca account list against AWS Organizations / Azure Tenants /
    GCP Organization root.
  • Use /orca-account-health to confirm Orca's view; cross-check
    externally for accounts that never made it to Orca.

"Connected but Empty"

⚠ <account> reports "connected" but Orca has zero assets and zero events.

This is the worst kind of blind spot — the integration looks healthy in
the dashboard but no data is actually flowing.

Likely causes:
  • IAM role attached but policy missing read permissions
  • Cross-account trust set up but external ID mismatch
  • Newly added account, scan not yet started
  • Region scope of the integration excludes the account's resources

Fix path:
  Orca Console → Cloud Accounts → <account> → Verify Permissions

MCP Tools Used

Primary Tools

Secondary Tools

Parameter Notes

• get_integration_configs_data takes no parameters — it's a full dump of configured sources. Filter client-side.
• discovery_search for "cloud audit events in the last 24 hours grouped by account" is the best signal that audit log ingestion is actually working — connection state alone is insufficient.
• discovery_search is capped at 50 results; for tenants with many accounts, narrow with "... in account <id>".

Implementation Notes

1. Two-source verification is the differentiator. get_integration_configs_data alone says "connected"; the discovery cross-check says "data is actually flowing." A "connected but empty" account is the worst blind spot and only the cross-check catches it.
2. Categorize the freshness, don't dump timestamps. Users need "Healthy / Sync Lag / Stale / Degraded / Offline" buckets, not raw "last_sync_at" values.
3. Always state the impact of each gap. Don't say "prod-eu-west has sync lag" — say "any alert triage in prod-eu-west is missing the last day's data". Connect the gap to the next action.
4. Pre-audit / pre-investigation framing matters. If the user invoked this before another skill ("before /orca-investigate"), end with an explicit "you can proceed" or "wait for re-sync" verdict.
5. External coverage drift (accounts in AWS Org but not in Orca) is out of scope for this skill — flag it in the edge case and recommend manual cross-check rather than pretend the MCP can detect it.
6. Link to other skills: this skill is a prerequisite for trusting the output of /orca-morning-briefing, /orca-investigate, /orca-compliance-gap, and /orca-cve-blast-radius. Mention this when appropriate.