- Home
- Skills
- Agents & Orchestration
- Work Order Writer — Delegation Contracts for Coding Agents
Works with the AI Agents you already use
Work Order Writer — Delegation Contracts for Coding Agents
Draft and audit verifiable work orders for coding agents to prevent scope creep and unverified claims.
$13
Work Order Writer — Delegation Contracts for Coding Agents
Example session with this skill installed
Write a work order for an agent to fix a timeout in the /api/process endpoint in src/server.ts. It should not change the database schema.
- Read your context and instructions
- Compiled the work order writer
I have drafted the work order for fixing the timeout in /api/process. It includes a measure-first baseline to capture the current latency and strict limits on schema modifications.
Work order: Fix timeout in /api/process endpoint
Objective
Optimize the processing logic in /api/process to resolve request timeouts.
Context
- Files:
src/server.ts,src/utils/processor.ts - Constraints: Node.js 20, Express, 5000ms timeout limit
- Known failures: Increasing the middleware timeout did not resolve the underlying compute bottleneck.
Skill to load
none
Measure first
Run curl -w "Time: %{time_total}\n" -X POST http://localhost:3000/api/process with the test-payload.json and record the duration.
Acceptance criteria
- AC1:
curlrequest withtest-payload.jsonreturns 200 OK in under 2000ms. - AC2:
npm run test:apiexits 0. - AC3: The request logs show the "Process complete" message with a timestamp diff under the limit.
Limits
Do NOT touch: src/db/schema.ts, do NOT add new npm dependencies, do NOT change the JSON structure of the API response.
Deliverable
A report with, per criterion: evidence (command output or diff), files changed, and anything left undone with the reason.
Next steps
- Copy this work order into the prompt for your coding agent.
- Once the agent provides a report, paste it here for a verification audit.
- Ensure the
test-payload.jsonis available in the root directory for the measure-first step.
Connects securely to your tools. The creator never sees your data.
About this skill
Delegating tasks to coding agents often results in "yes, it is done" reports that lack evidence or suffer from massive scope creep. This skill enforces a structured delegation contract to ensure every task is measurable, limited, and verifiable.
What it does
- Drafts work orders using a 7-field contract including objective, context, and explicit limits.
- Defines acceptance criteria as falsifiable commands or observable states rather than vague claims.
- Enforces baseline measurement by requiring a measure-first step before any code is modified.
- Audits agent reports by checking for command-output evidence and verifying no-touch violations.
- Standardizes handoffs between human developers and AI agents or between different agent roles.
How it works
- Provide the task description and file paths to the skill.
- Review the generated Work Order, ensuring the objective is exactly one sentence and criteria are runnable commands.
- Pass the Work Order to the executing agent.
- Provide the executing agent's report back to this skill for a line-by-line audit against the original criteria.
Frameworks & tools
This skill is language-agnostic and works with any terminal-based toolchain including npm, pytest, cargo, and npx. It is designed for use within agentic workflows in IDEs like Cursor or CLI agents like Claude Code.
Why this beats prompting it yourself
Writing tight, falsifiable constraints is tedious and easy to skip when in a hurry. This skill prevents "forever turns" and silent refactoring by forcing you to define "done" and "no-touch" zones before the agent starts burning tokens.
Use cases
- Drafting a bug fix ticket that requires a reproduction test before the fix.
- Delegating a specific refactor while strictly protecting public APIs from changes.
- Verifying a PR from an agent that claims success but provides no terminal output.
- Orchestrating a multi-agent team where the output of one must be the verified input of another.
Known limitations
Does not perform the actual coding; it acts as the architect and auditor. Effectiveness depends on the reviewer's ability to re-run at least one command locally.
Focuses on delegation and auditing only. Does not execute code changes. Requires the user or an agent to have terminal access to run specified commands.
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 20 days ago
Requires