agent-action-unit-tests
Apex test patterns for @InvocableMethod agent actions: per-branch coverage, bulk safety, deterministic assertions. NOT for UI/LWC testing or agent conversational quality scoring.
Best use case
agent-action-unit-tests is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Apex test patterns for @InvocableMethod agent actions: per-branch coverage, bulk safety, deterministic assertions. NOT for UI/LWC testing or agent conversational quality scoring.
Teams using agent-action-unit-tests should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/agent-action-unit-tests/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How agent-action-unit-tests Compares
| Feature / Agent | agent-action-unit-tests | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Apex test patterns for @InvocableMethod agent actions: per-branch coverage, bulk safety, deterministic assertions. NOT for UI/LWC testing or agent conversational quality scoring.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# Agent Action Unit Tests Agent actions are Apex classes with `@InvocableMethod`. They need the same testing rigor as any deployed Apex — plus two extras: every reason_code branch must be asserted, and the bulk-size contract (≤200 per invocation) must be verified. ## Recommended Workflow 1. List every reason_code the action can return; write one test per code. 2. Write a bulk test that submits 200 Request records and asserts 200 Response records come back — no Limits.getDMLStatements() >150 failures. 3. For callout actions, use `Test.setMock` with canned 200/4xx/5xx responses and assert the branch classification. 4. Assert on `Response.reason_code`, never on `Response.user_message` text (the text can change for UX; the code is the contract). 5. Run `sf apex run test -c` and confirm per-class coverage ≥85% AND every reason_code has an explicit assertion. ## Key Considerations - Coverage % alone is misleading; the requirement is per-reason-code assertion coverage. - Don't assert on user_message — it's UX copy; it will change. - Mock every callout; no real HTTP from tests. ## Worked Examples (see `references/examples.md`) - *Per-reason-code test matrix* — CloseCaseAction returns OK | VALIDATION_BLOCKED | UNKNOWN. - *Bulk-safety harness* — Agent batches 200 requests into one action invocation. ## Common Gotchas (see `references/gotchas.md`) - **Test.startTest/stopTest required for async** — Queueable from inside Invocable doesn't run in test without the boundary. - **@InvocableVariable default values** — Missing fields on Request become null, not default. - **Asserting on user_message strings** — UX change breaks 40 tests at once. ## Top LLM Anti-Patterns (full list in `references/llm-anti-patterns.md`) - Testing only the happy path with coverage padding. - Asserting on user_message text. - Skipping the bulk test because 'the agent only sends one at a time right now'. ## Official Sources Used - Agentforce Developer Guide — https://developer.salesforce.com/docs/einstein/genai/guide/agentforce.html - Einstein Trust Layer — https://help.salesforce.com/s/articleView?id=sf.generative_ai_trust_layer.htm - Invocable Actions (Apex) — https://developer.salesforce.com/docs/atlas.en-us.apexref.meta/apexref/apex_classes_invocable_action.htm - Agentforce Testing Center — https://help.salesforce.com/s/articleView?id=sf.agentforce_testing_center.htm
Related Skills
transaction-security-policies
Transaction Security policy creation and configuration: condition builder, enhanced policies, enforcement actions (block, MFA, notification, end session), real-time monitoring mode, and policy troubleshooting. NOT for Event Monitoring log analysis or Shield Event Monitoring setup (use event-monitoring). NOT for Apex testing or debug-log analysis.
omnistudio-remote-actions
Use when configuring, troubleshooting, or choosing between Remote Action types in OmniScript or FlexCard. Triggers: 'remote action', 'OmniScript action', 'IP action', 'Apex action element', 'VlocityOpenInterface2', 'Send/Response JSON Path'. NOT for Integration Procedure internal design (use integration-procedures) or generic Apex callout patterns (use apex integration skills).
lwc-quick-actions
Use when building a Lightning Web Component that runs from a record page quick-action button — either a screen action that renders UI in a modal or a headless action that invokes logic with no UI. Triggers: 'lwc quick action on record page', 'headless quick action no ui', 'closeactionscreenevent not working', 'how to pass recordid into quick action lwc', 'quick action vs flow action', 'quick action modal size'. NOT for Flow screen components — use `lwc-in-flow-screens` — and NOT for global actions without a record context or for list-view bulk actions that do not receive a single `recordId`.
flow-transactional-boundaries
Reason about when a Flow is inside the caller's transaction vs starts its own. Pick Before-Save vs After-Save vs Async Path vs Pause + Resume when transaction boundaries matter. Covers governor-limit sharing, DML sequencing, recoverability, and the exact semantics of each Flow entry point. NOT for choosing Flow vs Apex (use automation-selection.md). NOT for Flow-to-Flow invocation contracts (use subflows-and-reusability).
flow-transaction-finalizer-patterns
Use when a Flow needs to do work that must survive the triggering transaction — post-commit notifications, callouts, audit rows, or compensating actions. Covers Flow Transaction Control element, scheduled paths, Platform Event + finalizer escalation, and Apex Queueable finalizer bridging. Does NOT cover general Flow async decisions (see async-selection).
flow-record-save-order-interaction
Reason about how record-triggered flows interleave with the Salesforce Save Order (validation, before-save flows, before triggers, duplicate rules, after-save flows, workflow, after triggers, assignment, auto-response, escalation). Trigger keywords: save order, before-save flow, after-save flow, dml order, trigger vs flow order. Does NOT cover writing trigger handlers, approval process setup, or workflow rule migration.
flow-http-callout-action
Call external HTTP APIs directly from Flow using HTTP Callout actions (no Apex), handling auth, schema, and errors. NOT for complex Apex-based integration logic.
flow-action-framework
Use when designing or troubleshooting Salesforce Flow actions in Flow Builder: standard and core actions, the Apex action element for @InvocableMethod classes, how list-shaped inputs and outputs map at the Flow–Apex boundary, subflows, and choosing between declarative actions versus custom Apex versus packaged invocables. Triggers: 'Flow Apex action', 'add Apex to Flow', 'InvocableMethod in Flow', 'Flow action palette', 'map Flow variables to Apex invocable inputs'. NOT for authoring or testing Apex @InvocableMethod bodies (use the Apex invocable-methods skill), External Services or HTTP callout registration (use flow-external-services), OmniStudio action packs, or LWC screen component local actions.
github-actions-for-salesforce
Use this skill to set up, review, or troubleshoot GitHub Actions CI/CD pipelines for Salesforce using SFDX JWT Bearer Flow authentication, Apex test gates, and branch-conditional deployments. Trigger keywords: github actions, CI pipeline, jwt auth, sfdx ci, workflow yaml, github secrets, apex coverage threshold. NOT for other CI tools such as Jenkins, Copado, Bitbucket Pipelines, or Azure DevOps.
opportunity-pipeline-migration
Use when migrating historical Opportunity records into Salesforce — covering stage history recreation, amount and close date mapping, product line item load sequencing (Product2 → Price Book → OpportunityLineItem), team member (OpportunityTeamMember) and revenue split (OpportunitySplit) loading, and Pricebook2 assignment. NOT for generic multi-object data migration planning (use data-migration-planning). NOT for ongoing CRM sync or ETL pipeline architecture. NOT for opportunity record type or sales process configuration.
community-user-data-migration
Use this skill to migrate external community/portal user accounts at scale: bulk creating Experience Cloud users via Data Loader, migrating users between license types (Customer Community to Customer Community Plus, or to Partner Community), importing Customer Portal users into Experience Cloud, and resolving Contact/Account hierarchy prerequisites. Trigger keywords: migrate community users, import external users Experience Cloud, bulk create portal users, move users between license types, migrate Customer Community to Partner Community. NOT for internal user data migration. NOT for general data migration (see data/bulk-data-migration). NOT for configuring Experience Cloud sites or profiles from scratch.
community-analytics-data
Use when analyzing Experience Cloud site analytics including login metrics, member engagement, page view tracking, and content performance. Triggers: Experience Cloud site analytics, community member engagement data, portal login tracking, page view reports community, GA4 Experience Cloud integration. NOT for CRM Analytics or Tableau CRM. NOT for internal Salesforce reporting on standard CRM objects.