nexus-gateway: Skill for Claude Code

.claude/skills/wire-capability-probe/SKILL.md

wire-capability-probe is a skill for Claude Code from AlphaBitCore/nexus-gateway. It costs 0 tokens per session (3,262 once invoked), scanned B, original, Apache-2.0.

A procedure for finding out whether an external service supports the data format your gateway sends, whether the gateway sends the wrong format, or whether the selected model lacks support. A gateway is software that translates and forwards requests between your application and a provider.

In plain words
What is it for?
Use it when a provider rejects a request because of its content type, field, input mode, or URL format. It guides tests through the gateway, directly in the gateway's format, and directly in the provider's format.
Why use it?
An error from an external service does not always identify the real fault. Comparing gateway and direct requests helps avoid recording the wrong capability or changing the wrong part of the system.

Skill for Claude Code

Written for Claude Code: installed under .claude/. Also seen: reads .claude/ paths.

This is AlphaBitCore/nexus-gateway's own configuration. It tells Claude Code how to work on nexus-gateway itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything nexus-gateway configures →

Reuse

Borrowing it

Nothing to install: this file belongs to AlphaBitCore/nexus-gateway. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/AlphaBitCore/nexus-gateway/main/.claude/skills/wire-capability-probe/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/AlphaBitCore/nexus-gateway

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for wire-capability-probe

README.md
[![agentmods](https://agentmods.dev/badge/skills/alphabitcore/nexus-gateway/wire-capability-probe.svg)](https://agentmods.dev/skills/alphabitcore/nexus-gateway/wire-capability-probe)
Your own site
<a href="https://agentmods.dev/skills/alphabitcore/nexus-gateway/wire-capability-probe"><img src="https://agentmods.dev/badge/skills/alphabitcore/nexus-gateway/wire-capability-probe.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,262 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 2 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00000 $0.03262
Opus 5 $0.00000 $0.01631
Sonnet 5 $0.00000 $0.00652
Haiku 4.5 $0.00000 $0.00326

Measured 7d ago against content hash 320e919f5dda, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade B, and why

wire-capability-probe scanned grade B with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Sends data to an external URLmediumData exfiltration

A POST to an outside endpoint may be telemetry or may be exfiltration; either way the mod talks to somewhere, and you should know where.

Then `curl --data @<file>` at the vendor. This is what closes the loop: shape assertions passed

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

Then `curl --data @<file>` at the vendor. This is what closes the loop: shape assertions passed
.claude/skills/wire-capability-probe/SKILL.md · 290 lines

How it starts

The opening of the file, as written. The whole thing — 290 lines — stays where its author put it; the contents beside it link to each section on GitHub.

wire-capability-probe

An upstream refusal is not an answer. It is one of three different answers wearing the same clothes:

Verdict What it means What fixes it
The wire cannot This API has no shape for what the caller sent Refuse in our words, naming what and why; record the capability
We send the wrong form The wire carries it, under a different name or in a different place A codec mapping
The model cannot The wire carries it; this model does not Routing / catalog, not the codec

Reading the refusal message tells you which one it is only when you are lucky. This skill is the procedure for when you are not, and its rules were each bought by a wrong verdict.


The core procedure: three probes, never one

For every refusal, run all three. The third is the one that changes verdicts.

1. THROUGH THE GATEWAY   — reproduce it. This is the symptom.
2. DIRECT, OUR SHAPE     — same bytes, straight to the vendor, no gateway.
                           Same refusal?  → not a mistranslation.
                           Different?     → the gateway is the problem. Stop here.
3. DIRECT, VENDOR SHAPE  — the form the vendor's own docs describe.
                           200?           → "we send the wrong form". Codec mapping.
                           Refused?       → the wire genuinely cannot.

Worked example, and the reason this skill exists. Cohere answered 422 unrecognized content type 'file' for a document. Probe 2 reproduced it exactly, which looked like confirmation. Probe 3 sent the same document in Cohere's documented top-level documents array: 200, and the model answered from the document's contents. The verdict flipped from "Cohere cannot take documents" — which was about to be written into the model catalog — to "our codec has an unused mapping".

Three refusals probed the same day gave three different verdicts:

Vendor Our shape, direct Vendor's documented shape Verdict
Cohere 422 unrecognized content type top-level documents → 200, answered our gap
Moonshot 400, identical to via-gateway GET /v1/files → 200; upload-then-reference, no inline form no inline form — an upload capability, not a mapping
DeepSeek 400 unknown variant nothing documented genuinely unsupported

Read the full file on GitHub · 290 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 7d ago First seen · 290 lines · 0 tokens per session scan B 320e919f5dda

Subscribe to this mod's changes

wire-capability-probe is a skill published in the GitHub repository AlphaBitCore/nexus-gateway (23 stars, last pushed 5d ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 3,262 tokens. A static security scan graded it B with 2 findings (sends data to an external url, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.