publish-hotline

Publish local programs as callable hotlines on the delegated-execution network with declared contracts.

1|Updated Mar 19, 2026
One-click install
npx skills add https://github.com/hejiajiudeeyu/delegated-execution-workspace --skill publish-hotline-hejiajiudeeyu
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: publish-hotline
Source: https://github.com/hejiajiudeeyu/delegated-execution-workspace/tree/main/.claude/skills/publish-hotline
Command: npx skills add https://github.com/hejiajiudeeyu/delegated-execution-workspace --skill publish-hotline-hejiajiudeeyu

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Publishing a local script, CLI, model, or service as a callable hotline fails silently when the contract is incomplete: the hotline registers, the catalogue looks fine, and callers only discover the problem when calls fail. This Skill walks through writing the worker, declaring a complete contract, registering, submitting, approving, and verifying the hotline is genuinely callable. ## Core Features & Use Cases - Worker authoring guidance: Defines the process-adapter worker pattern — print the contract JSON on --contract, otherwise read one JSON task on stdin and write one JSON result to stdout. - Publication gate compliance: Explains the local submit-review gate (caller-facing descriptions on every input field) and the approval gate (CONTRACT_HOTLINE_INCOMPLETE checks for schemas, examples, and stated limitations). - End-to-end verification: Verifies callability via contract-check, the platform catalogue endpoint, and the agent-facing view, including delivery integrity checks (verified vs unchecked). - Production trap catalog: Documents shipped failure modes such as stale registration drafts, dropped service tiers before client 0.1.23, missing contract_role on output files, and port conflicts with running devices. - Use Case: You wrote a PDF-parsing script and want agents on the network to call it. Use this Skill to declare its input/output schemas and attachment requirements, register it with delexec-ops add-hotline, pass both approval gates, and confirm agents see the platform-published contract. ## Quick Start Use the publish-hotline skill to turn my local script into a callable hotline, register it, and verify agents can call it.

Frequently Asked Questions about publish-hotline

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I publish a local script as a callable hotline?▼

Write a process-adapter worker that prints its contract JSON on --contract and otherwise reads a JSON task on stdin. Then register it with delexec-ops add-hotline, run submit-review, and have the operator approve both the hotline and its responder device.

Why does submit-review fail with HOTLINE_INPUT_GUIDANCE_REQUIRED?▼

Every property in input_schema.properties must carry a caller-facing description, and the old template phrases like "source text" are on a denylist. Add a real description to each field, then delete the stale registration draft with remove-hotline before re-registering, since add-hotline reuses the old draft.

Why is my approved hotline invisible in the catalogue?▼

The catalogue publishes an entry only when the hotline's responder device is also approved and routable. Approving only the hotline leaves it frozen with catalog_visibility "hidden" and a 404 on /v2/hotlines/:id, so approve the pending responder too.

How do I verify a hotline is actually callable by agents?▼

Run delexec-ops responder contract-check, fetch the published contract from the platform catalogue, and check the agent-facing view where contract_source must be platform_catalog. Then place one real call and confirm the platform reports delivery_integrity as verified.

Why does my hotline show the wrong service_tier or execution budget?▼

Clients before 0.1.23 read service_tier, execution_budget_s, and acceptance window from the worker but dropped them before writing the registration draft, so a quick hotline published as standard. Repair requires re-registration and fresh approval because the corrected declaration changes the digest.

What are the limitations of local testing for priced hotlines?▼

The local stack has BILLING_ENFORCEMENT unset, so fixed-price calls settle as none and the caller is never charged. A green local run proves the call works, not the price; use tools/paid-call-e2e.mjs, which runs its own Postgres with enforcement enabled.