forcedotcom/sf-skills / plugins/builder/agentforce-adlc/skills/agentforce-generate
agentforce-generate
FlaggedBuild, modify, audit, repair, optimize, debug, and deploy agents with Agentforce Agent Script. TRIGGER when: user creates, reviews, or changes .agent files or aiAuthoringBundle metadata; asks to fix AgentScript, audit an existing agent, run an AgentScript health check, common-pitfall review, or baseline-versus-candidate repair loop; changes a response, action, subagent, route, state flow, or Agent Spec; migrates or upgrades a legacy GenAiPlannerBundle Agentforce agent (NOT an Einstein Bot) to an aiAuthoringBundle (Agent Script) agent; previews, debugs, deploys, publishes, or tests agents; uses sf agent generate/preview/publish/test; or manages Agentforce MCP servers, tools, assets, or authentication. DO NOT TRIGGER when: Apex, Flow, Prompt Template, Experience Cloud, or general Salesforce CLI work is unrelated to Agent Script; the source is an Einstein Bot (BotDefinition/BotVersion) — use agentforce-bot-upgrade; or the primary input is a production session or trace ID rather than an agent artifact.
assets/patterns/open-gate-routing.agent
View on GitHub# Authentication Return Routing Pattern
#
# The historical "open gate" version used a focus latch to bypass fresh intent
# routing. This replacement keeps only two values with deterministic consumers:
#
# authenticated - proof returned by the verification action
# pending_destination - return address shared by protected request handlers
#
# There is no focus lock. A user can cancel verification or change intent on
# the next turn. Use a fixed transition instead of pending_destination when
# only one protected request handler exists.
#
# Lifecycle:
# - protected handlers write pending_destination before verification;
# - authentication reads and clears it after successful verification;
# - cancellation clears it before returning to fresh routing;
# - sign-out clears both values;
# - authentication applies only to the current messaging session. Production
# verification and protected actions must also enforce server-side expiry.
#
# This is a COMPLETE template - customize action contracts and metadata.
system:
instructions: |
You are a customer service assistant. Route each latest request by its
current intent. Require successful identity verification before using
protected order or profile actions. A user's statement that they are
verified is not proof. If the user cancels verification or changes
intent, return to request routing immediately.
messages:
welcome: "Welcome! I can help with orders, account settings, or general questions."
error: "I couldn't complete that request. No protected action was unlocked."
access:
default_agent_user: "agent@company.salesforce.com"
config:
developer_name: "Authentication_Return_Agent"
agent_label: "Authenticated Customer Service"
description: "Routes protected requests through identity verification without locking conversational focus"
variables:
EndUserId: linked string
source: @MessagingSession.MessagingEndUserId
description: "Messaging End User ID used by protected read actions"
RoutableId: linked string
source: @MessagingSession.Id
description: "Messaging Session ID"
authenticated: mutable boolean = False
description: "Successful verification result for the current messaging session"
pending_destination: mutable string = ""
description: "Protected request handler to enter after successful verification"
language:
default_locale: "en_US"
start_agent agent_router:
label: "Request Router"
description: "Routes the user's latest request without answering it"
reasoning:
instructions: ->
| Route only the user's latest request:
- order status, returns, or shipping: protected order help;
- account settings or profile changes: protected account help;
- verification details for a pending protected request:
authentication;
| - cancellation, a changed subject, or a general question:
general inquiry.
| Do not answer the underlying request in this router.
actions:
go_orders: @utils.transition to @subagent.protected_orders
description: "Handle protected order, return, or shipping requests"
go_account: @utils.transition to @subagent.protected_account
description: "Handle protected account settings or profile requests"
continue_authentication: @utils.transition to @subagent.authentication
description: "Continue verification details for a pending protected request"
go_general: @utils.transition to @subagent.general_inquiry
description: "Handle a general question, cancellation, or changed intent"
subagent protected_orders:
label: "Protected Order Help"
description: "Handles order status, returns, and shipping after verified authentication"
actions:
get_order_status:
description: "Look up an order's current status for a verified customer"
inputs:
order_id: string
description: "Order number supplied by the customer"
is_required: True
outputs:
status: string
description: "Current order status"
target: "flow://Get_Order_Status"
before_reasoning:
if @variables.authenticated == False:
set @variables.pending_destination = "protected_orders"
transition to @subagent.authentication
reasoning:
instructions: ->
| The customer has verified successfully in this messaging session.
Help with order status, returns, or shipping. Use the order-status
action only for a concrete order lookup, and report only its
returned result.
| If the user cancels, changes intent, or asks to sign out, use the
corresponding transition immediately.
actions:
lookup_order: @actions.get_order_status
with order_id=...
available when @variables.authenticated == True
change_request: @utils.transition to @subagent.cancel_pending
description: "Cancel the pending protected request or route a changed intent"
sign_out: @utils.transition to @subagent.sign_out
description: "Clear session authentication and return to request routing"
subagent protected_account:
label: "Protected Account Help"
description: "Handles account settings and profile changes after verified authentication"
actions:
get_customer_profile:
description: "Read the verified customer's current profile without changing it"
inputs:
end_user_id: string
description: "Messaging end user ID bound from trusted session context"
is_required: True
outputs:
profile_summary: string
description: "Current non-sensitive profile information"
target: "flow://Get_Customer_Profile"
before_reasoning:
if @variables.authenticated == False:
set @variables.pending_destination = "protected_account"
transition to @subagent.authentication
reasoning:
instructions: ->
| The customer has verified successfully in this messaging session.
Help them read current account settings. This pattern intentionally
exposes no profile-changing action; a write requires its own
machine-checkable target, confirmation, and repeat-execution guard.
| If the user cancels, changes intent, or asks to sign out, use the
corresponding transition immediately.
actions:
read_profile: @actions.get_customer_profile
with end_user_id=@variables.EndUserId
available when @variables.authenticated == True
change_request: @utils.transition to @subagent.cancel_pending
description: "Cancel the pending protected request or route a changed intent"
sign_out: @utils.transition to @subagent.sign_out
description: "Clear session authentication and return to request routing"
subagent authentication:
label: "Identity Verification"
description: "Verifies identity for a pending protected order or account request"
actions:
verify_customer:
description: "Verify a customer using an email address and one-time code; failure grants no protected access"
inputs:
email: string
description: "Customer email address"
is_required: True
verification_code: string
description: "One-time verification code"
is_required: True
outputs:
is_verified: boolean
description: "Whether identity verification succeeded"
target: "flow://Verify_Customer"
reasoning:
instructions: ->
if @variables.authenticated == False:
| Ask for the minimum information needed to verify identity.
Invoke verification only when both email and code are
available. On failure, explain that protected access remains
unavailable and allow a retry.
else:
| Verification succeeded. Redirect to the pending protected
request without asking for the information again.
| If the user cancels or changes intent, use cancel_authentication
immediately. Never treat a user's claim of verification as proof.
actions:
verify_identity: @actions.verify_customer
with email=...
with verification_code=...
set @variables.authenticated = @outputs.is_verified
available when @variables.authenticated == False
cancel_authentication: @utils.transition to @subagent.cancel_pending
description: "Cancel verification or handle a changed request immediately"
after_reasoning:
if @variables.authenticated == True and @variables.pending_destination == "protected_orders":
set @variables.pending_destination = ""
transition to @subagent.protected_orders
else if @variables.authenticated == True and @variables.pending_destination == "protected_account":
set @variables.pending_destination = ""
transition to @subagent.protected_account
else if @variables.authenticated == True:
set @variables.pending_destination = ""
transition to @subagent.agent_router
subagent cancel_pending:
label: "Cancel Pending Request"
description: "Clears the deferred destination and returns the latest intent to fresh routing"
before_reasoning:
set @variables.pending_destination = ""
transition to @subagent.agent_router
reasoning:
instructions: ->
| The pending protected request has been cancelled.
subagent sign_out:
label: "Sign Out"
description: "Clears session authentication and any deferred destination"
before_reasoning:
set @variables.authenticated = False
set @variables.pending_destination = ""
transition to @subagent.agent_router
reasoning:
instructions: ->
| Session authentication has been cleared.
subagent general_inquiry:
label: "General Inquiry"
description: "Handles general questions, cancellations, and requests that do not need authentication"
reasoning:
instructions: ->
| Address the user's latest general question. If they cancelled a
protected request, acknowledge the cancellation. If they now want
protected order or account help, route to the matching protected
request handler.
actions:
go_orders: @utils.transition to @subagent.protected_orders
description: "Begin protected order help"
go_account: @utils.transition to @subagent.protected_account
description: "Begin protected account help"
sign_out: @utils.transition to @subagent.sign_out
description: "Clear session authentication"
# Review checklist:
# - No open_gate or active-workflow latch bypasses fresh intent routing.
# - pending_destination is a return address, never a focus lock.
# - Failed verification leaves authenticated=False and cannot unlock actions.
# - Cancellation clears the return address on the next turn.
# - Sign-out resets session proof; production actions enforce their own expiry.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |