More screenshots
Works with the AI tools you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsAgent Output Actionability Gate
Decide whether an AI-agent output is safe to hand off using an evidence ledger, contract checks, approval gates, and a clear PASS, HOLD, or ESCALATE verdict.
$12
Agent Output Actionability Gate
Example session with this skill installed
Our support agent drafted this action for review:
“Refund request completed for customer Maya Chen, order 8841. Refund $189.00 to the original card and email the customer that it is done.”
The downstream adapter expects order_id, amount, currency, reason, and approval_id. Refunds above $100 require a support-lead approval. The agent output includes no currency, no reason, no approval ID, and no payment-provider receipt. The customer record is not available. Review the output before it reaches the refund tool.
- Read your context and instructions
- Compiled the agent output actionability
VERDICT
HOLD
HANDOFF SUMMARY
The proposed handoff would send a refund request for order 8841 to a payment tool and then send a customer-facing completion message. The action is financial and externally visible. It is not reversible by the agent’s wording alone, and the output claims completion even though no provider receipt was supplied.
OUTPUT LEDGER
Completion claim — “Refund request completed.” Status: UNKNOWN. No payment-provider receipt or verified state was supplied. Impact: the customer message must not say the refund is complete.
Target — order 8841. Status: SUPPLIED, but customer identity and order ownership are not independently verified. Impact: confirm the order in the authorized system before execution.
Amount — 189.00. Status: SUPPLIED, but currency is missing. Impact: the adapter contract is incomplete and the amount cannot be safely interpreted.
Reason — missing. Status: UNKNOWN. Impact: required by the downstream contract.
Approval ID — missing. Status: BLOCKER because refunds above $100 require support-lead approval.
FINDINGS
BLOCKER — Missing approval ID. A refund above the policy threshold cannot reach the tool without a named approval record.
HIGH — Missing currency and reason. The adapter contract is not satisfied.
HIGH — Unsupported completion claim. The text describes a successful refund without a receipt or state verification.
MEDIUM — Customer-facing message is premature. Draft a pending-review message only after the approval and tool result exist.
APPROVAL AND ESCALATION
A support lead must approve the refund and confirm the authorized order target. A payment operator or verified adapter must return a receipt before anyone tells the customer that the refund completed.
REPAIR CHECKLIST
- Add
currencyand a concrete refundreason. - Attach the support-lead
approval_idfor the $189.00 request. - Verify order 8841 and the authorized customer identity in the payment system.
- Replace “completed” with “awaiting approval” until a provider receipt confirms the state.
- Record the tool request ID, provider receipt, final state, and customer-message approval.
VERIFICATION
The output is ready only when the adapter validates all required fields, the approval ID is present, the provider receipt confirms the final refund state, and the customer message matches that verified state.
Connects securely to your tools. The creator never sees your data.
What you get
About this skill
Agent Output Actionability Gate reviews the moment where an AI response becomes someone else’s input: a human approval, another agent, a workflow step, or a tool-call adapter. A fluent answer can still be incomplete, out of scope, unsupported, malformed, or falsely claim that an external action succeeded.
Provide the original task, the proposed output, the next consumer, and any schema, policy, approval rule, evidence, or execution receipt you have. The skill reconstructs the handoff, separates confirmed facts from model assertions, checks required fields and scope, and identifies whether the response is advisory, preparatory, requested for execution, or actually verified.
You receive an output ledger, severity-ranked findings, missing-proof requests, approval and escalation requirements, a repair checklist, and one verdict: PASS, PASS WITH CONDITIONS, HOLD, ESCALATE, or INSUFFICIENT CONTRACT. It does not execute tools, send messages, approve transactions, verify live systems, or invent missing policies.
How to install
Works the same in every agent - Claude, Cursor, Codex, Copilot and 20+ more.
- 1
Download the ZIP
Free skills download straight away. Paid skills unlock right after purchase.
- 2
Unzip into your skills folder
Every agent reads skills from one folder on your machine. Drop the unzipped folder in there.
- 3
Ask your agent to use it
Restart the agent if it was already running. It picks the skill up automatically - no config needed.
Skills folder by agent
Click the path to copy it. Create the folder if it does not exist yet.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
30-day refund
Not a fit? Get your money back
Trust & safety
Security scanned
Verified clean 18 days ago
- Passed all security checks, Safe to install