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/critical-input-collection.agent
View on GitHub# Critical Input Collection Pattern
# Ensures reliable capture of critical data before proceeding with workflows
#
# ★ Problem This Solves:
# - LLM sends empty JSON to actions
# - LLM sends wrong field names (_id instead of account_id)
# - LLM extracts wrong values from conversation
# - Agent proceeds without required data, causing crashes
#
# ★ Root Cause:
# Slot filling (`with account_id=...`) relies on LLM inference, which is
# probabilistic and can fail in subtle ways. Critical inputs need
# deterministic collection with validation.
#
# ★ Key Patterns:
# 1. First-Interaction Collection - Tell LLM its PRIMARY GOAL
# 2. Variable Setter Action - Dedicated action to capture + validate + store
# 3. Single-Use Pattern - `available when @var == ""` disables after use
# 4. Null Guard Pattern - Block ALL downstream actions until input is valid
# 5. Explicit Action References - Guide LLM with `{!@actions.x}` in instructions
#
# ★ When To Use This Pattern:
# - Workflows that REQUIRE a specific ID (account, order, case, etc.)
# - When slot filling has been unreliable
# - When invalid input causes downstream failures
# - Multi-subagent agents where input must persist across transitions
#
# This is a COMPLETE template - customize for your use case
system:
instructions: "You are a research assistant. Your PRIMARY GOAL is to collect the account ID before performing any research. Never proceed without a validated account ID."
messages:
welcome: "Hello! I can help you research account information. Please provide the account ID to get started."
error: "I encountered an issue. Let me try again."
access:
default_agent_user: "agent@company.salesforce.com"
config:
developer_name: "Account_Research_Agent"
agent_label: "Account Research Agent"
description: "Demonstrates critical input collection patterns for reliable slot filling"
variables:
# Standard linked variables (required)
EndUserId: linked string
source: @MessagingSession.MessagingEndUserId
description: "Messaging End User ID"
RoutableId: linked string
source: @MessagingSession.Id
description: "Messaging Session ID"
ContactId: linked string
source: @MessagingEndUser.ContactId
description: "Contact ID"
# ★ CRITICAL INPUT - The value that MUST be collected before proceeding
account_id: mutable string = ""
description: "The Salesforce Account ID (18-char, starts with 001) - MUST be collected and validated before any research"
# ★ VALIDATION STATE - Track whether input has been validated
account_id_validated: mutable boolean = False
description: "Whether account_id has been validated as a real Salesforce ID"
# ★ COLLECTION ATTEMPTS - Track retries to prevent infinite loops
collection_attempts: mutable number = 0
description: "Number of times we've tried to collect account_id"
# Research results
account_name: mutable string = ""
description: "Name of the account (from lookup)"
research_summary: mutable string = ""
description: "Research summary for the account"
language:
default_locale: "en_US"
additional_locales: ""
all_additional_locales: False
# ★ ENTRY POINT: Input Collection Subagent
# This subagent's ONLY job is to collect and validate the critical input
start_agent input_collector:
label: "Input Collector"
description: "Collects and validates the account ID before allowing research"
actions:
# ★ PATTERN 2: Variable Setter Action
# Dedicated action that captures, validates, and stores the input
capture_account_id:
description: "Captures and validates the Salesforce Account ID from user input"
inputs:
account_id: string
description: "The 18-character Salesforce Account ID provided by the user (format: starts with '001')"
is_required: True
outputs:
validated_account_id: string
description: "The validated account ID (empty if invalid)"
is_valid: boolean
description: "True if the account ID is valid and exists"
account_name: string
description: "Name of the account if found"
error_message: string
description: "Error message if validation failed"
target: "flow://Validate_Account_Id"
require_user_confirmation: False
include_in_progress_indicator: True
progress_indicator_message: "Validating account ID..."
reasoning:
# ★ PATTERN 1: First-Interaction Collection
# Tell the LLM its PRIMARY GOAL is to collect the input
instructions: ->
| YOUR PRIMARY GOAL: Collect the account ID from the user.
CURRENT STATE:
if @variables.account_id == "":
| ⚠️ Account ID NOT YET COLLECTED
ASK the user for their Salesforce Account ID.
The ID should be 18 characters and start with "001".
Example prompt: "Please provide the Account ID you'd like me to research."
if @variables.account_id != "" and @variables.account_id_validated == False:
| 🔄 Account ID received: {!@variables.account_id}
Attempting validation...
Use {!@actions.capture_account_id} to validate this ID.
if @variables.account_id_validated == True:
| ✅ Account ID validated: {!@variables.account_id}
Account Name: {!@variables.account_name}
Continue to the research workflow.
if @variables.collection_attempts > 2 and @variables.account_id_validated == False:
| ❌ Multiple validation failures.
Ask the user to verify their Account ID is correct.
Suggest they check Salesforce for the correct ID format.
actions:
# ★ PATTERN 3: Single-Use Pattern
# Action becomes unavailable once account_id is successfully captured
validate_id: @actions.capture_account_id
with account_id=...
set @variables.account_id = @outputs.validated_account_id
set @variables.account_id_validated = @outputs.is_valid
set @variables.account_name = @outputs.account_name
set @variables.collection_attempts = @variables.collection_attempts + 1
# ★ SINGLE-USE: Unavailable once validated
available when @variables.account_id_validated == False
# ★ PATTERN 4: Null Guard Pattern
# Transition ONLY available when input is validated
go_research: @utils.transition to @subagent.research
available when @variables.account_id_validated == True
# ★ RESEARCH SUBAGENT: Only accessible after input is validated
subagent research:
label: "Account Research"
description: "Performs research on the validated account"
actions:
# Research action that USES the validated account_id
research_account:
description: "Performs detailed research on the account"
inputs:
account_id: string
description: "The validated account ID"
is_required: True
outputs:
research_summary: string
description: "Summary of research findings"
signal_summary: string
description: "Key signals detected"
target: "flow://Research_Account"
include_in_progress_indicator: True
progress_indicator_message: "Researching account..."
reasoning:
instructions: ->
| Researching account: {!@variables.account_name} ({!@variables.account_id})
if @variables.research_summary == "":
| Use {!@actions.research_account} to analyze this account.
if @variables.research_summary != "":
| Research complete!
{!@variables.research_summary}
actions:
# ★ PATTERN 4: Null Guard - Uses variable binding, NOT slot filling
# This ensures the validated ID is used, not LLM re-extraction
do_research: @actions.research_account
with account_id=@variables.account_id
set @variables.research_summary = @outputs.research_summary
# ★ GUARD: Only available if we have validated ID (defensive)
available when @variables.account_id_validated == True
available when @variables.research_summary == ""
# Return to collector for new account
new_account: @utils.transition to @subagent.input_collector
available when @variables.research_summary != ""
# ═══════════════════════════════════════════════════════════════════════════
# ★ INSIGHTS: Critical Input Collection
# ═══════════════════════════════════════════════════════════════════════════
#
# WHY SLOT FILLING FAILS:
# The `...` syntax relies on LLM to extract values from conversation.
# This is probabilistic and can fail when:
# - User provides ID in unexpected format
# - Multiple IDs mentioned in conversation
# - LLM abbreviates or modifies field names
# - Context window is cluttered with other data
#
# THE SOLUTION: DETERMINISTIC COLLECTION
# Instead of `with account_id=...` on EVERY action:
# 1. Collect ONCE with a dedicated setter action
# 2. Validate BEFORE storing
# 3. Store in variable: `set @variables.account_id = @outputs.validated_id`
# 4. Use variable binding: `with account_id=@variables.account_id`
# 5. Guard all actions: `available when @variables.account_id_validated == True`
#
# PATTERN SUMMARY:
#
# ┌─────────────────────────────────────────────────────────────────┐
# │ CRITICAL INPUT COLLECTION FLOW │
# ├─────────────────────────────────────────────────────────────────┤
# │ │
# │ User provides ID ────► Setter Action (slot fills once) │
# │ │ │
# │ ▼ │
# │ Validation │
# │ │ │
# │ ┌────────┴────────┐ │
# │ ▼ ▼ │
# │ Valid Invalid │
# │ │ │ │
# │ set @variables.account_id Ask again │
# │ set @variables.validated=True (limit retries) │
# │ │ │
# │ ▼ │
# │ All downstream actions use │
# │ @variables.account_id (NOT ...) │
# │ │
# └─────────────────────────────────────────────────────────────────┘
#
# ═══════════════════════════════════════════════════════════════════════════
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 |