- Home
- Skills
- Agents & Orchestration
- Agentforce Release Passport
Works with the AI Agents you already use
Agentforce Release Passport
Know if your Agentforce agent is safe to promote before you deploy. Checks dependencies in the target org, tests the agent against your acceptance criteria, and gives a PASS or BLOCK verdict with an evidence report.
$19
Agentforce Release Passport
Example session with this skill installed
Check whether Support_Assistant can be promoted from my-dev to my-qa. Runs per golden: 3. Threshold: 0.8.
Acceptance criteria
AC-1: A refund request routes to the resolution subagent.
AC-2: The agent creates a real case and tells the customer the case number.
- Read your context and instructions
- Compiled the agentforce release passport
- Generated the document
Verdict: BLOCK ⛔
Blocked for two reasons. First, AC-2 is not met: in 9 of 9 runs the agent never created a case, and in some runs it told the customer "a ticket has been logged" when no Case exists. Second, my-qa is missing 8 things the agent needs.
Verdict conditions
| # | Condition | Result |
|---|---|---|
| 1 | Scan has no errors | ✅ |
| 2 | No travel gaps | ⛔ 8 gaps |
| 3 | Stamp exists | ✅ |
| 4 | Stamp not stale | ✅ fingerprint matches |
| 5 | All goldens at or above threshold (0.8) | ⛔ G6, G7, G8 at 0/3 |
| 6 | No regressions against baseline | ✅ first stamp |
Acceptance criteria coverage
| Criterion | Result |
|---|---|
| AC-1: refund routes to resolution | 12/12 runs passed |
| AC-2: real case + case number | 0/9 on the happy path ⛔ |
Root cause (AC-2)
create_support_case is only available when kb_article_found is true. The knowledge base search returned no article in any run, so the agent could only escalate.
Travel gaps (in deploy order)
- ApexClass:TicketClassifier: missing
- Flow:Lookup_Order, Search_Help_Articles, Create_Support_Case, Escalate_To_Human: missing
- Agent metadata: include in deployment
- PermissionSet:Support_Agent_Access, Agent_Data_Access: missing
- Einstein Agent User: create manually
Scoped deploy commands are in the report. Nothing was deployed.
Warnings
- The agent claims actions it never took: "is being escalated" when escalate_case didn't run. The platform's groundedness check marked these replies GROUNDED.
- escalate_case receives an empty case_id and unexpanded template text.
Full passport saved to .passport/passport-Support_Assistant-my-qa.md
agentforce-release-passport.pdf
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
An Agentforce agent that works in Dev can still fail after promotion. A Flow it calls never made it to QA, or a small change quietly sends "I want a refund" to the wrong topic. Agentforce Release Passport catches both before the agent moves.
WHAT IT DOES
- Scan: maps everything the agent depends on: topics/subagents, actions, Flows, Apex classes and prompt templates. Works with Agent Script and classic agents.
- Travel check: confirms every dependency exists in the target org, including inactive Flows, missing permission sets and the agent user. Gives you scoped deploy commands in the right order.
- Goldens: turns your user story's acceptance criteria into test conversations, including edge cases and negative tests. You approve them before anything runs.
- Stamp: runs each test several times (LLM agents don't answer the same way every time) and records a pass rate per test, compared against previous runs to catch regressions.
- Verify: a clear PASS or BLOCK, with a release passport your release manager can read without context.
WHAT MAKES IT DIFFERENT
- It decides whether an agent is safe to ship, not just how to build it.
- It checks the real target org, not just your branch.
- It traces every test back to an acceptance criterion.
- It catches agents claiming actions they never took, like "a ticket has been logged" when no Case exists. In testing, it caught replies that Salesforce's own groundedness check marked as grounded.
SAFE BY DESIGN
- Read-only on the target org. Never deploys, activates or deletes without your explicit yes.
- Tests run only in non-production orgs.
- It never weakens a test to get a PASS. Any change to tests needs your approval and is recorded in the report.
REQUIREMENTS
- Salesforce CLI (sf) with the agent commands
- Source and target orgs authorised with sf org login
- Works in Claude Code, Cursor and other agents that support skills
- Tested on Windows with Agent Script agents
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Requires local Salesforce CLI auth. Testing limited to non-production orgs. Windows requires PowerShell for reliable CLI interaction. Metadata fingerprints become stale if agent files change.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 9 days ago
Permissions
Salesforce CLI (sf)