salesforce-flow-automation-review-skill

Salesforce Flow Automation Review Skill

Purpose

This skill reviews Salesforce Flow XML, validation rules, approval processes, and record-triggered automation for correctness, safety, and maintainability. It flags patterns that cause data corruption, runaway governor-limit consumption, incorrect execution order, and logic errors. It does not access live orgs and does not authorize automation activation.

When to use

• A Flow is being reviewed before activation in production.
• A record-triggered automation has caused unexpected behavior and must be analyzed.
• An automation inventory contains mixed automation types on the same object.
• An approval process or validation rule needs pre-release safety review.

When not to use

• Apex code is the primary subject — use salesforce-apex-lwc-code-review-skill.
• Metadata quality (field definitions, layouts) — use salesforce-metadata-review-skill.
• Live automation activation proposal — use salesforce-live-change-approval-protocol.
• Full org automation inventory — use salesforce-org-assessment-skill.

Minimum payload (required inputs)

• Flow XML (or equivalent exported Flow definition) or automation description.
• Automation type: record-triggered, screen, scheduled, autolaunched, or subflow.
• Object(s) the automation acts on.
• Whether other automation types (Apex triggers, Process Builder

, validation rules) are active on the same object.

Workflow

1. Recursion detection

• Identify flows that update the same object that triggered them.
• Check for recursion prevention mechanisms: $Flow.CurrentStage, custom boolean

field as recursion guard, or explicit entry condition excluding re-triggered records.

• Flag flows that could recursively fire without a guard.

2. Bypass flags

• Flag flows that check custom bypass permissions or hierarchy settings to skip

execution (e.g., Bypass_All_Flows__c custom field or permission set bypass).

• Determine whether bypass is documented, scoped, and auditable.
• Flag ungoverned bypass: bypass that has no associated approval process or

documentation.

3. Null handling

• Review Decision elements for null-check branches.
• Flag formulas or conditions that will throw errors on null values.
• Flag flows that access lookup fields without a null guard on the lookup itself.
• Flag Get Records elements whose output is used without a null check after

the element.

4. Fault paths

• Verify that every element that can fail (DML, invocable actions, callouts,

platform events) has a connected Fault path.

• Flag flows where a Fault path is missing entirely.
• Flag Fault paths that silently swallow errors (no logging, no notification).

5. Before-save vs after-save misuse

• Record-triggered flows operating before-save: flag if they make DML operations

(not allowed in before-save context).

• Record-triggered flows operating after-save: flag if they could be moved to

before-save for efficiency (field updates only, no related-record operations).

• Flag after-save flows that cause cascading DML triggering Apex triggers.

6. Hardcoded recipients and IDs

• Flag hardcoded email addresses in Send Email elements.
• Flag hardcoded user IDs or group IDs in assignment or notification elements.
• Recommend: Custom Metadata, Custom Labels, or hierarchy settings instead.

7. Mixed automation on same object

• Identify if Process Builder

(deprecated), Apex triggers, and Flows are all active on the same object.

• Flag execution order risks: trigger order between Apex and Flows is not always

deterministic; flag if both Fire at the same DML moment.

• Recommend: migration plan to consolidate to a single automation layer.

8. Approval process review

• Flag approval processes without email templates (hardcoded notification text).
• Flag approval processes where the final approver is a specific user ID.
• Flag approval processes with no rejection branch behavior defined.
• Flag approval processes on objects also governed by Flow with duplicate logic.

9. Validation rule review

• Flag validation rules with no descriptive error message.
• Flag validation rules that reference fields not on the current object without

a VLOOKUP — this can cause silent failures.

• Flag validation rules that fire on all events when they should be

conditioned on specific record types or entry points.

Evidence requirements

• Exported Flow XML or detailed automation description; no credentials, no session tokens.
• Object name(s) and list of other active automation on the same object.
• If the automation accesses external systems (callouts), note the endpoint category.

Output format

flow_automation_findings:
  recursion_risks:
    - finding: [description]
      severity: Critical | High | Medium | Low
      evidence: [element name or condition]
      recommendation: [brief]
  bypass_flag_risks: [same structure]
  null_handling_risks: [same structure]
  fault_path_gaps: [same structure]
  before_after_save_issues: [same structure]
  hardcoded_values: [same structure]
  mixed_automation_risks: [same structure]
  approval_process_findings: [same structure]
  validation_rule_findings: [same structure]

summary:
  total_findings: [count]
  critical_count: [count]
  high_count: [count]
escalation_gates_fired: [from salesforce-risk-taxonomy, or "none"]
assumptions: [list]
missing_evidence: [what would improve the review]

Redaction rules

• Never request secrets, credentials, OAuth tokens, refresh tokens, session IDs, MFA seeds, customer PII.
• Sanitize org IDs, user IDs (replace with placeholders) before sharing in outputs.
• Do not repeat hardcoded email addresses or user IDs from Flow XML in output; describe pattern and location.

Privilege / data handling rules

• Automation review is logic-level only; do not carry record data from Flow outputs.
• Findings involving data exposure patterns escalate to salesforce-permission-model-review-skill.

Handoff rules

• Hands off to: salesforce-apex-lwc-code-review-skill (if Apex triggers are involved),

salesforce-permission-model-review-skill (if permission bypass is found), salesforce-org-assessment-skill (if mixed automation is systemic).

• If escalation gate fires: salesforce-case-capsule with escalation_required = true.
• Required handoff fields: matter_id, automation_type, object_name, escalation_gates_fired,

critical_count.

Audit log fields

• matter_id, skill_id, skill_version, invoked_by, input_hash, evidence_quality, output_verdict, escalation_fired, timestamp

Stop conditions

• Flow XML contains live credentials, callout secrets, or PII — stop and ask for sanitized version.
• Recursion without guard detected in a production-active flow — output Critical finding and recommend immediate deactivation review with human authorization.
• Ungoverned bypass flag grants admin-level access to non-admin users — escalate to salesforce-permission-model-review-skill.

Security notes

• Read-only static review of exported automation definitions; never requests live org access.
• Hardcoded recipient values must not be repeated in output.
• Automation activation or deactivation decisions require human-authorized change management.
• Process Builder deprecation status should be verified against current Salesforce release notes.