- Home
- Skills
- Agents & Orchestration
- Readable Work-Intent Contract for Orchestrators and Fleets
Readable Work-Intent Contract for Orchestrators and Fleets
When a worker machine runs jobs from somewhere else, its owner should be able to read what each job is for.
$29
Works with the AI Agents you already use
Readable Work-Intent Contract for Orchestrators and Fleets
Example session with this skill installed
Validate this job description before I send it to the render fleet: {"schemaVersion":1,"title":"task-4471","project":"Forest cartoon","purpose":"Produce the forest scene for the final cartoon","stage":"Rendering","expectedOutput":"120 image frames"}
- Read your context and instructions
- Compiled the readable work-intent contract
Summary
Rejected: title is an identifier, not a description.
Detail
node -ewithvalidateWorkIntent-> Error: Provide a readable title (8-160 characters); identifiers and empty placeholders are not descriptions.python3 -I scripts/work_intent.py validate job.json-> INVALID with the same reason (exit 1); both implementations agree.
Fix
Replace the title with what the job is, for example "Render cartoon episode 1, scene 4, frames 120-239". Re-validated: VALID. A worker showing this queue will display that sentence; a legacy worker would have shown "Description unavailable - producer update required" instead of the id.
Connects securely to your tools. The creator never sees your data.
About this skill
When a worker machine runs jobs from somewhere else, its owner should be able to read what each job is for. Left alone, producers send task-4471, a UUID, or "unknown". This contract makes the description a validated object and makes both of your implementations agree on what passes.
The contract (reference/contract.md): schemaVersion 1 plus title (8-160), project (2-80), purpose (12-600), stage (2-80), expectedOutput (6-400) and optional parentTitle. A field is readable only when it is trimmed, within its UTF-16 length range, well-formed, contains a letter, carries no control, BOM, zero-width or bidirectional characters, is not an identifier (digit runs, 16+ hex, UUIDs, or those prefixed with session/task/job/lease/pid/thread/run/build) and is not a placeholder. Unknown fields such as command or permissions fail with their names listed: this object describes work, it never authorizes it.
Parsing: 8192-byte cap before parsing; duplicate top-level keys rejected even with equal values and even when the second key is spelled with JSON unicode escapes, because JSON.parse silently keeps the last one; returned object frozen.
Legacy projection: legacyTitle() keeps a readable label or your own policy's explicit "withheld" marker and replaces everything else with "Description unavailable - producer update required", so a queue never shows an id.
Two implementations, one fixtures file: work-intent.mjs (Node 20+) and work_intent.py (Python 3.9+, with a validate <file> CLI). 29 Node tests including a cross-check that runs the 5-test Python suite on the same fixtures. No packages, no network.
Lexical checks only; does not verify truthfulness of prose. Placeholder lists are English-centric. Requires manual porting of new rules to both Node and Python files.
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 18 days ago
Requires
Permissions
Pure validation library plus a CLI that reads one JSON file. No network, no environment variables, no writes.