Held Order
by
Held order decisions for ecommerce teams: work out why each order is stuck and what it is worth, then ship it, investigate it, or cancel it.
Orders get held for a dozen different reasons and then sit in one undifferentiated queue: a fraud score above threshold, an address the carrier cannot validate, a payment awaiting capture, a SKU that went out of stock between order and pick. They are treated as one problem because they look like one list. The cost is asymmetric and invisible: every day a legitimate order sits held is a customer forming an opinion, and every fraudulent order released is a chargeback plus the goods. This skill works out why each order is stuck and what it is worth, then ships, investigates, or cancels.
What It Does
- Separates the reasons orders are held rather than treating the queue as one thing.
- Reads the customer's history against the fraud signal that triggered the hold.
- Prices the cash sitting in the queue and how long it has been there.
- Flags the address problem that is a typo rather than a risk.
- Escalates only what genuinely needs a person to look at it.
- Ships, investigates, or cancels each order, with the reason recorded.
Use Cases
- Release the legitimate orders sitting behind a blunt fraud rule.
- Find the orders held for three days on a correctable address typo.
- See the cash stuck in the held queue right now.
- Give a reviewer only the orders that genuinely need judgment.
Example Session
What is in the held queue, and what should happen to each order?
Forty-one orders, $14,600. Thirty-three can be released today.
- Reason held | Orders | Value | Action | Why
- Fraud score, repeat customer | 19 | $6,800 | Ship | 3+ prior orders, no chargebacks
- Address unvalidated, likely typo | 14 | $4,100 | Ship, corrected | Single character variance
- Fraud score, first order, mismatch | 5 | $2,900 | Investigate | Billing and shipping differ
- Payment not captured | 3 | $800 | Cancel | 9 days, no response
Nineteen of these are repeat customers with clean histories caught by a score that cannot see their history. They have been waiting an average of two days for nothing. The address group is fourteen typos the validator rejected on one character. Five orders genuinely need a person, which is a queue someone can actually work through today.
Why This Beats Prompting It Yourself
A Fraud Score vs The Customer Behind It: The Hidden Risk
A generic prompt trusts the score, because the score is the signal the system provides. A score computed on the transaction alone cannot see that this customer has ordered four times, kept everything and never filed a chargeback. Holding that order is a false positive that costs a real customer and no fraud was prevented.
When You Still Need The Five That Need A Person
The value of triage is what it leaves for a human. Thirty-six of these forty-one orders have a defensible automatic answer, and the five that do not share a specific pattern: a first order with a billing and shipping mismatch. Handing a reviewer five orders with the reason attached is a different task from handing them a queue of forty-one.
Known Limitations
Address correction is proposed where the variance is a single character and never applied silently on a larger difference, because shipping to the wrong corrected address is worse than the hold. Fraud judgment uses your own chargeback history, so a store without one relies more heavily on the provider's score and says so. It sets the action and does not release, cancel or capture payment in your platform.
Install It
- Download the ZIP.
It is free and there is no account to create. - Unzip it into your agent's skills folder.
Claude Code reads ~/.claude/skills/, which is hidden by default: the command in the folder block opens it. Other agents scan their own directory, so drop the same folder there instead. - Ask your agent to use it.
Restart the agent if it was already running, then it picks the skill up with no config.
skills folder
Copy
~/.claude/skills/held-order/
SKILL.md
references/hold-reasons.md
queue/
# one file per run, with the action and reason
# macOS: create the folder and open it in Finder
mkdir -p ~/.claude/skills && open ~/.claude/skills
# Windows: paste in the Explorer address bar
%USERPROFILE%\.claude\skills\Frequently asked questions
How is this different from my fraud tool's recommendations?
A fraud tool scores the transaction. This reads the score against the customer's order and chargeback history with you, separates fraud holds from address and payment holds, and prices the cash sitting in the queue. Most of what it releases was never a fraud question.
Which agents does the skill run in?
Any agent that supports the open SKILL.md format: Claude Code, Cursor, Codex CLI, GitHub Copilot, Gemini CLI, Manus, Grok and others load it unmodified. The format is portable, the location is not. Each agent scans its own skills directory, so you drop the same folder into whichever one yours uses. There is no config file to edit and no API key to provision.
What does it need connected to work?
Held orders with their hold reasons, customer order and chargeback history, and address validation results. Without chargeback history it is more conservative and labels the output accordingly.
Is it safe to release orders a fraud tool flagged?
For a customer with several clean prior orders, yes, and that is the bulk of what sits in most queues. The skill never releases a first order with a billing and shipping mismatch, which is the pattern where the score is usually right.
Why cancel rather than keep chasing an uncaptured payment?
Because after nine days without a response the order is not going to complete, and the stock has been reserved that whole time. The threshold is configurable, and the skill states how long each order has been waiting so the cut point is visible.
What is included with the skill?
The skill itself, the hold reason reference, the queue decision format, and the release thresholds. It is free and security scanned. We re-publish the ZIP when the thresholds change, so download it again if your release rate shifts.
Ship, investigate or cancel is a margin and risk call.
Polar joins order value, customer history and fraud signals in one semantic layer, so the action you take weighs both sides.
<script type="application/ld+json">{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"How is this different from my fraud tool's recommendations?","acceptedAnswer":{"@type":"Answer","text":"A fraud tool scores the transaction. This reads the score against the customer's order and chargeback history with you, separates fraud holds from address and payment holds, and prices the cash sitting in the queue. Most of what it releases was never a fraud question."}},{"@type":"Question","name":"Which agents does the skill run in?","acceptedAnswer":{"@type":"Answer","text":"Any agent that supports the open SKILL.md format: Claude Code, Cursor, Codex CLI, GitHub Copilot, Gemini CLI, Manus, Grok and others load it unmodified. The format is portable, the location is not. Each agent scans its own skills directory, so you drop the same folder into whichever one yours uses. There is no config file to edit and no API key to provision."}},{"@type":"Question","name":"What does it need connected to work?","acceptedAnswer":{"@type":"Answer","text":"Held orders with their hold reasons, customer order and chargeback history, and address validation results. Without chargeback history it is more conservative and labels the output accordingly."}},{"@type":"Question","name":"Is it safe to release orders a fraud tool flagged?","acceptedAnswer":{"@type":"Answer","text":"For a customer with several clean prior orders, yes, and that is the bulk of what sits in most queues. The skill never releases a first order with a billing and shipping mismatch, which is the pattern where the score is usually right."}},{"@type":"Question","name":"Why cancel rather than keep chasing an uncaptured payment?","acceptedAnswer":{"@type":"Answer","text":"Because after nine days without a response the order is not going to complete, and the stock has been reserved that whole time. The threshold is configurable, and the skill states how long each order has been waiting so the cut point is visible."}},{"@type":"Question","name":"What is included with the skill?","acceptedAnswer":{"@type":"Answer","text":"The skill itself, the hold reason reference, the queue decision format, and the release thresholds. It is free and security scanned. We re-publish the ZIP when the thresholds change, so download it again if your release rate shifts."}}]}</script>