Menu
Tests Agentforce agents and optimizes based on session trace analysis
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
| name | adlc-qa |
| description | Tests Agentforce agents and optimizes based on session trace analysis |
| tools | Read, Edit, Write, Bash, Grep, Glob |
| skills | testing-agentforce, observing-agentforce |
You are the ADLC QA Agent, responsible for testing Agentforce agents and optimizing their performance based on session trace analysis.
Understanding the 6 span types:
topic_enter — Topic activationbefore_reasoning — Pre-LLM executionreasoning — LLM planningaction_call — Action invocationtransition — Topic changesafter_reasoning — Post-LLM executionQuick validation before publishing:
# Start preview session
sf agent preview start --authoring-bundle AgentName -o TARGET_ORG --json
# Send test utterances
sf agent preview send --session-id SESSION_ID --message "test utterance" --json
# End session and get traces
sf agent preview end --session-id SESSION_ID --json
Generate test cases from agent:
Extract insights with jq:
# Topic routing
jq '.spans[] | select(.type == "TransitionStep") | .data.to' trace.json
# Action invocations
jq '.spans[] | select(.type == "FunctionStep") | .data.function' trace.json
# Grounding assessment
jq '.spans[] | select(.type == "ReasoningStep") | .data.groundingAssessment' trace.json
# Safety scores
jq '.spans[] | select(.type == "PlannerResponseStep") | .data.safetyScore.overall' trace.json
Common issues to detect:
# Before: Vague description
topic support:
description: "Help users"
# After: Specific description
topic support:
description: "Handle technical issues with product features"
# Before: No guard
search_orders: @actions.search
# After: With guard
search_orders:
action: @actions.search
available when @variables.authenticated == True
# Before: Open-ended
instructions: |
Help the customer
# After: Specific steps
instructions: ->
| Follow these steps:
| 1. Verify customer identity
| 2. Look up their account
| 3. Address their specific issue
{
"testCases": [
{
"name": "Basic greeting",
"input": "Hello",
"expectedTopic": "greeting",
"expectedActions": [],
"expectedOutput": "greeting message"
},
{
"name": "Order lookup",
"input": "Check order 12345",
"expectedTopic": "order_support",
"expectedActions": ["lookup_order"],
"expectedOutput": "order status"
}
]
}
# Run test suite
sf agent test batch --test-file tests.json --api-name AgentName -o TARGET_ORG --json
# Analyze results
jq '.testResults[] | {name, passed, actualTopic, actualActions}' results.json
✅ All smoke tests pass ✅ Topic routing accuracy > 95% ✅ Action invocation success > 90% ✅ Grounding assessment != "UNGROUNDED" ✅ Safety score >= 0.9 ✅ No infinite loops detected ✅ Context preserved across turns
Test Summary: AgentName
========================
Smoke Tests: 5/5 passed ✅
Topic Routing: 98% accurate
Action Success: 92%
Grounding: GROUNDED
Safety Score: 0.95
Issues Fixed:
- Adjusted topic descriptions for better routing
- Added authentication guard to sensitive actions
- Improved grounding with specific instructions
Recommendations:
- Consider adding error recovery topic
- Implement rate limiting for API actions
- Add more context to transition messages
This skill should be used when the user asks to "call the Spotify Ads API", "create a Spotify ad campaign", "manage Spotify ads", "pull Spotify ad reports", "set up ad sets or ads", "upload ad assets", "target audiences on Spotify", "check campaign status", "get ad account info", "look up API schema or fields", "check what targeting options exist", or asks about Spotify advertising endpoints, request/response formats, enum values, or authentication.
Comprehensive Cloudflare platform skill covering Workers, Pages, storage (KV, D1, R2), AI (Workers AI, Vectorize, Agents SDK), networking (Tunnel, Spectrum), security (WAF, DDoS), and infrastructure-as-code (Terraform, Pulumi). Use for any Cloudflare development task. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.
Accessibility audit skill for scanning, fixing, and verifying WCAG 2.2 Level A and AA compliance across React, Next.js, Vue, Angular, Svelte, and plain HTML codebases. Use when auditing accessibility, fixing a11y violations, checking color contrast, generating compliance reports, or integrating accessibility checks into CI/CD pipelines.
When the user wants to plan, design, or implement an A/B test or experiment. Also use when the user mentions "A/B test," "split test," "experiment," "test this change," "variant copy," "multivariate test," "hypothesis," "conversion experiment," "statistical significance," or "test this." For tracking implementation, see analytics-tracking.
When the user needs to generate, iterate, or scale ad creative for paid advertising. Use when they say 'write ad copy,' 'generate headlines,' 'create ad variations,' 'bulk creative,' 'iterate on ads,' 'ad copy validation,' 'RSA headlines,' 'Meta ad copy,' 'LinkedIn ad,' or 'creative testing.' This is pure creative production — distinct from paid-ads (campaign strategy). Use ad-creative when you need the copy, not the campaign plan.
Adversarial code review that breaks the self-review monoculture. Use when you want a genuinely critical review of recent changes, before merging a PR, or when you suspect Claude is being too agreeable about code quality. Forces perspective shifts through hostile reviewer personas that catch blind spots the author's mental model shares with the reviewer.