- Home
- Skills
- Prompt & Skill Engineering
- Tool Schema Designer
Works with the AI Agents you already use
Tool Schema Designer
Repair tool schemas so models pick the right tool and fill it correctly, proven with a before/after routing test.
$15
Tool Schema Designer
Example session with this skill installed
Our agent keeps calling refund_order when the user only asks where their order is, and it invents order IDs. Here are the tool definitions and 20 bad calls.
- Read your context and instructions
- Compiled the tool schema designer
Tool schema review: orders server
Verdict
refund_order and get_order_status share the verb "order" and both accept an unconstrained order_id string, so the model guesses. Names and constraints are the cause, not the model.
Findings
T1 — Misrouted "where is my order"
Observed: 7 of 20 bad calls hit refund_order with a fabricated ID.
Cause: descriptions describe implementation; neither says when not to use it.
Change: get_order_status_by_id and refund_order_after_confirmation, each with a "when not to use" line pointing at the other.
Expected effect: misroutes 7/20 → 1/20.
T2 — Invented order IDs
Cause: order_id is an unconstrained string with no example.
Change: pattern ^A-\d{4}$, example A-1042, error text names the format and the lookup alternative.
Expected effect: invented IDs 4/20 → 0/20.
T3 — Refund without confirmation
Change: required confirm: true with a description naming the amount, plus destructiveHint annotation.
Error contract
Every failure returns: what was wrong, the valid shape, and the next tool to try. No stack traces.
Before/after
| Case | Before | After |
|---|---|---|
| "where is my order" | refund_order | get_order_status_by_id |
| "refund A-55" | invented ID | get_order_status_by_id then confirm |
Rollout
Renaming is breaking for cached clients; ship as v2 and keep v1 for two weeks.
Connects securely to your tools. The creator never sees your data.
About this skill
Tool schema design and repair: names, descriptions and constraints that make models route correctly, proven with a before-and-after test.
Use this skill when:
- the model picks the wrong tool
- tool calls have bad arguments
- improve function definitions
- too many tools confuse the agent
Who it's for: Agent developers with misused tools.
Example request
The agent keeps refunding instead of answering order status.
What the example returns
- Two tools share the verb order
- Renamed, described, constrained
- Misroutes 7/24 to 1/24
In the package: five files — SKILL.md with the trigger and workflow, README.md, references/METHOD.md (the full method), references/CHECKLIST.md (the pass/fail gate) and examples/EXAMPLE.md (a worked run you can copy).
Works with Claude Code, Cursor, Codex CLI and any agent that reads SKILL.md. Plain markdown: no API key, no account, no network access.

Requires call logs for behavioral diagnosis. Structural review only without logs. Schema name or parameter changes are breaking changes for clients.
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 any payment.
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 2 days ago
Permissions
Reviews definitions and logs you provide. No network calls, no credentials.