Atlas / Skills / antevo-ch / plugins

pluginsCAUTION

skills/antevo-ch/plugins

Antevo for Claude — Executive, Trademark, Wealth and Mandates connections with 25 skills. Public where it can be, OAuth where it's personal.

Verdict
CAUTION
Grade
C
Trust score
79 /100
Version
—
Hosts
5 documented
License
NOASSERTION
Stars
0
01

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

Antevo brings the world around your wealth into the assistant you already use. Four connections — the Executive Brief, trademark registers, your own household and your firm's client book — and 27 skills that know how to read them: dated, attributed, and honest about what they cannot see. Install once in Claude, or point any MCP client at an address.

[](https://github.com/ANTEVO-CH/plugins/actions/workflows/validate.yml) [](https://registry.modelcontextprotocol.io/v0/servers?search=ch.antevo) [](#skills) [](https://www.npmjs.com/package/@antevo/cli) [](https://antevo.ch)

Why Antevo

  • A named desk, not a web search. Answers come from Antevo's published Executive Brief and its dated archive, from the trademark registers themselves, and — when you sign in — from your own record. The assistant says where a read came from and when.
  • Public where it can be, permissioned where it matters. Executive and trademark screening need no account and reach no personal data. Wealth and Mandates sign in over OAuth 2.1; your account's own permissions apply.
  • Intelligence, not advice. The skills hold the line: no invented probabilities, no "clear to use", no buy or sell dressed up as a signal.

Who this is for

  • **Families and private in
Read from source at commit 8d1fc49b292aOBSERVED · 2026-10-10
02

Host compatibility

What the documentation claims. We have not run a compatibility test.

HostStatusNotes
claude-codementioned
claude-desktopmentioned
cursormentioned
gemini-climentioned
windsurfmentioned
03

What it tells the agent

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: decode
description: >-
  Explain what a market move actually means, in plain language, using the Antevo
  desk's own decode. Use when the user asks "what does this mean", "explain
  this", "why does that matter", "I don't follow markets, what's going on",
  "help me understand", "break this down", "what should I take from this", or
  asks about a market event in a way that shows they want understanding rather
  than a briefing. Also for "teach me how to read this". Universal — no account,
  no portfolio access. If they want the professional editorial read instead, use
  world-brief. Read-only.
compatibility: >-
  Requires the Antevo Executive connector (public, no sign-in). If the tools are
  not available, tell the user to add https://api.antevo.ch/mcp/executive/mcp,
  then stop.
---

# Antevo — Decode

For the reader who wants to *understand* the move, not be briefed on it.

## The one thing that makes this skill work

**The desk has already written the plain-language version. Use it. Do not write
your own.**

Simplifying a financial claim is where meaning quietly changes — a hedge gets
dropped, a probability becomes a certainty, "the odds fell" becomes "it won't
happen". The brief carries an authored jargon-free rendering of the whole thing,
so there is no reason to paraphrase and every reason not to.

## Step 1 — Pull it

```
get_executive_brief(sections="plain_language")
```

That's the whole skill's input — about 2KB rather than the full 24KB brief.

| Field | What it is |
|-------|------------|
| `headline` | The move in one plain sentence |
| `decode.tape_says` | What happened, read quickly — the surface story |
| `decode.lesson` | The transferable principle — how to read *this kind of* move |
| `decode.capital_does` | How an allocator separates the questions the headline blends |
| `signal_read` | The probability layer, without jargon |
| `theme_reads` | One line per theme — Gold, Energy, Equities, Rates |

Add `sections="plain_language,regime"` if they asked what kind of market this is.

## Step 2 — Build the arc

`decode` is already a teaching sequence. Follow it in order — the order is the
pedagogy:

1. **`tape_says`** — the fast read. Start where the reader already is.
2. **`lesson`** — the principle. This is the part they keep after they've
   forgotten the news.
3. **`capital_does`** — what a professional does differently with the same
   headline. This is the payoff.

Then `theme_reads` for anything they asked about specifically.

## Step 3 — Pitch it right

- **Assume no jargon, but never talk down.** The desk's own copy sets the level;
  match it.
- **Lead with the concrete.** A named waterway, a named metal, a date — not
  "risk assets".
- **One idea per paragraph.** If a sentence has two clauses joined by "while",
  it's probably two paragraphs.
- **Answer what they asked.** If they asked about gold, `theme_reads` has a gold
  line — start there and use the decode to explain why it moved that way.

## Step 4 — Output

```
# {headline}

{tape_says — what happened, in one short paragraph}

## What's actually going on
{the distinction the headline blends — from capital_does}

## The thing worth remembering
{lesson — stated as a principle they can reuse}

## Where it showed up
- **{theme}** — {read}

---
*From the Antevo Executive Brief, {date}. Editorial market intelligence — not
investment advice.*
```

Drop any section they didn't need. A good decode of one question beats a full
template.

## Guardrails
- **Quote the desk's framing; don't invent your own simplification.** That's the
  point of the skill.
- **Keep the hedges.** If the copy says "looked close, then", it does not say
  "failed". Certainty is the easiest thing to add by accident and the most
  damaging.
- **A lesson is not a prediction.** "Check whether the distant dates moved too"
  teaches a habit; it does not say what happens next.
- **No personal advice, ever** — this reader is the most likely to take a
  general remark as one. If they ask what they should do, say plainly that this
  is universal commentary, and point them to the Antevo Wealth connector for
  anything about their own holdings.
- If `plain_language` is missing for a date, use `world-brief` instead and say
  you're giving the professional read — don't improvise a simplified one.
04

Trust audit

CAUTIONgrade C · trust 79/100 Install with care. The audit found things worth knowing before you trust its output.

LayerWhat it checksResult
L0Provenance & inventoryPASS
L1Static analysis of the codePASS
L2Instruction surface (what it tells the agent)FAIL
L3Class-specific surfacePASS
L4Behavioural (sandbox)SKIPPED

What the source does

Filesystem
none-observed
Network
none-observed
Shell
none-observed
Dependencies
pinned
Secrets in source
none-found

Findings (1)

HIGHPrompt injection · prompt.hide_from_user · CWE-94, CWE-1427
plugins/antevo-wealth/skills/advisor-meeting-prep/SKILL.md:113
- **Questions, not verdicts.** Never tell the user to buy, sell, move money or
Why it matters. asks the agent to act without the user's knowledge

Gates applied: instruction_override, no_behavioural_pass.

Audited 2026-10-10 · audit v0.4.1 · source sha 8d1fc49b292afull audit observations/trust-audit/skill/antevo-ch__plugins.json · Report an issue / request a re-scan
05

Audit history

Every audit this skill has had.

DateSourceVerdictGradeScoreChange
2026-10-108d1fc49b292aCAUTIONC79first audit
06

Questions

What does the plugins skill do?

Antevo for Claude — Executive, Trademark, Wealth and Mandates connections with 25 skills. Public where it can be, OAuth where it's personal.

Is plugins safe to install?

With care. The audit graded it C (79/100) and found 1 thing worth knowing before you trust this skill, listed below with the exact line each was found on.

What can plugins access on my machine?

The audit observed no filesystem, network or shell use at all in its source.

Which assistants does plugins work with?

Its documentation mentions claude-code, claude-desktop, cursor, gemini-cli and windsurf. That is what the text claims, not a compatibility test we ran.

How current is this page?

The grade is for one exact copy of the source (8d1fc49b292a), read on 2026-10-10. The repository is watched, and a new audit runs when it changes — this is the first audit.

Advertisement