Salesforce release readiness skill

Salesforce Release Readiness Skill

Purpose

This skill evaluates a Salesforce release for deployment readiness. It produces a structured readiness assessment covering sandbox strategy, source tracking, package diff, destructive changes, test coverage, post-deploy steps, rollback plan, communications plan, and approval matrix. It does not execute deployments, access live orgs, or authorize production changes.

When to use

• A Salesforce release is being prepared and readiness must be assessed.
• A change set or DX package deployment is planned and the checklist must be verified.
• A destructiveChanges.xml must be reviewed before a production deploy.
• A release retrospective revealed a gap in the pre-deploy process.

When not to use

• Live production mutation approval at execution time — use salesforce-live-change-approval-protocol.
• Code quality review — use salesforce-apex-lwc-code-review-skill.
• Automation logic review — use salesforce-flow-automation-review-skill.
• Metadata quality review — use salesforce-metadata-review-skill.

Minimum payload (required inputs)

• Release description: what is being deployed (features, fixes, metadata types).
• Target environment: sandbox → sandbox, sandbox → production (use placeholder, not real org ID).
• Deployment method: change set, SFDX/SF CLI deploy, CI/CD pipeline, ISV package install.
• Available artifacts: package diff, destructiveChanges.xml (if applicable), test run output.
• Rollback plan (or note that it is missing).

Workflow

1. Sandbox refresh strategy

• Verify that testing occurred in a sandbox refreshed from production within a defined window (configurable threshold, e.g., 30 days).
• Flag: testing only in a developer sandbox without a full-copy or partial-copy refresh for complex data-dependent features.
• Flag: sandbox refresh done after feature development began (data divergence risk).

2. Source tracking state

• Review source tracking status: are there untracked local changes?
• Flag: uncommitted changes in tracked source that are not part of the release.
• Flag: metadata retrieved outside source tracking that may overwrite tracked changes.
• Verify: scratch org ( Salesforce DX) or sandbox state matches the deployment package.

3. Package version diff

• Review the list of metadata components in the deployment package.
• Flag: metadata types that are unexpectedly included (scope creep).
• Flag: metadata components that are expected but missing from the package.
• Flag: large diff (> configurable threshold components) without a phased deployment plan.

4. Destructive changes review

• If destructiveChanges.xml is present:
  • List all components marked for deletion.
  • Flag: deletion of custom fields on objects with active sharing rules or automation referencing those fields.
  • Flag: deletion of Apex classes referenced by other classes not in the package.
  • Flag: deletion without a data migration plan for fields containing data.
  • Flag: deletion of permission sets that are actively assigned.
• If destructiveChanges.xml is absent but deletions are expected: flag missing destructive changes manifest.

5. Test coverage threshold

• Verify that Apex test coverage meets the org threshold (default 75%; flag if org has a higher custom threshold).
• Flag: test classes with SeeAllData=true in the test run.
• Flag: test run results showing failures (zero failures required for production deploy).
• Flag: test coverage calculated only on deployed classes, not org-wide (insufficient).
• Recommend: run all tests (RunAllTestsInOrg) for production deployments.

6. Post-deploy steps

• Verify that a post-deploy step list exists and is documented.
• Flag: manual configuration steps not captured in the package (e.g., custom metadata values, permission set assignments, named credential setup).
• Flag: post-deploy steps that require production data access without a documented data-access control.

7. Rollback plan

• Verify that a rollback plan exists and is actionable.
• Flag: rollback plan that is "redeploy the previous package" without a version reference.
• Flag: rollback plan for destructive changes (field/object deletion is irreversible without data backup).
• Recommend: document rollback estimated time and who authorizes rollback.

8. Communications plan

• Verify that a user communications plan exists for user-visible changes.
• Flag: changes to page layouts, record types, or permissions without a comms plan.
• Flag: scheduled deployments during business hours without maintenance window notification.

9. Approval matrix

• Verify that the approval matrix is documented: who approved the release, at what level (developer, tech lead, architect, change advisory board).
• Flag: production deployments without documented architecture or change board approval for significant changes.
• Flag: emergency deployments without a documented fast-track approval path.

Evidence requirements

• Release description, deployment method, and target environment.
• Package component list or diff.
• Test run output (or note that it is unavailable).
• Rollback plan (or note that it is missing).

Output format

release_readiness_assessment:
  sandbox_refresh:
    status: ready | at-risk | not-ready
    findings: [list]
  source_tracking:
    status: ready | at-risk | not-ready
    findings: [list]
  package_version_diff:
    status: ready | at-risk | not-ready
    findings: [list]
  destructive_changes:
    status: ready | at-risk | not-ready | not-applicable
    findings: [list]
  test_coverage:
    status: ready | at-risk | not-ready
    findings: [list]
  post_deploy_steps:
    status: ready | at-risk | not-ready
    findings: [list]
  rollback_plan:
    status: ready | at-risk | not-ready
    findings: [list]
  communications_plan:
    status: ready | at-risk | not-ready
    findings: [list]
  approval_matrix:
    status: ready | at-risk | not-ready
    findings: [list]

overall_readiness: ready | at-risk | not-ready
blocking_issues: [list — items that must be resolved before deploy]
non_blocking_issues: [list — recommendations for improvement]
escalation_gates_fired: [from salesforce-risk-taxonomy, or "none"]
assumptions: [list]
missing_evidence: [what would improve the assessment]

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 include real usernames or approver identities; use role references.

Privilege / data handling rules

• Release artifacts must not contain production data samples.
• Destructive changes review involving regulated-data fields escalates to compliance review.

Handoff rules

• Hands off to: salesforce-live-change-approval-protocol (if overall_readiness = ready and production deploy is next), salesforce-apex-lwc-code-review-skill (if test coverage gaps need code-level review), salesforce-case-capsule (if escalation_gates_fired).
• Required handoff fields: matter_id, overall_readiness, blocking_issues, escalation_gates_fired.

Audit log fields

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

Stop conditions

• Destructive changes include field deletion on objects with unreviewed data — stop and flag irreversible-deploy gate.
• Test run shows failures — overall_readiness = not-ready; do not proceed.
• Rollback plan is absent and deployment includes irreversible changes — stop and require rollback plan before assessment can complete.

Security notes

• Read-only advisory assessment; never executes deployments or requests live org access.
• Destructive changes are irreversible; always require human-authorized rollback plan documentation.
• Approval matrix verification is a governance control; its absence is a high-severity finding.