---
name: argus-multi-agent-workflow
description: Coordinate machine-readable Argus fact retrieval with structured, neutral handoffs.
---

# Argus Data Layer Handoffs

## Customer-Side Installation

Set `ARGUS_BASE_URL`, run `argus auth login`, and use the live Agent Tool Registry as the runtime authority. Approved unattended clients may read `ARGUS_MACHINE_API_KEY`; do not pass credentials through a remote shell.

Authentication is independent for each participant; discover capabilities with `tool:agent_tool_registry`. Share only returned structured facts and metadata when permissions and license terms allow it.

Use explicit subjects, symbols, filters, and request `as_of`; validate returned `effective_as_of` as response metadata. Use `tool:evidence_bundle_export` for a reproducible structured handoff. Include audit identifiers and coded coverage gaps.

Discover with `argus tool-registry` and apply binding `parameter_aliases` and `parameter_encodings`. CLI/MCP list arguments are comma-separated strings; REST lists are JSON arrays. Read live OpenAPI or MCP `list_tools().inputSchema` for executable shapes. Business responses use `success`/`result`/`audit_id`; registry discovery is a raw snapshot. With OAuth omit caller/institution fields. Check MCP `isError` before decoding structured content and resolve real production subjects rather than copying example identifiers.

Argus does not orchestrate Agents, create narratives, recommend actions, execute strategies or let ordinary Agents control account security. Personal lists require explicit owner authorization. Credentials are never shared between participants.

## OAuth tool grants

Discovery does not grant execution rights. Check `permission_requirements.default_oauth_access` and `discovery_group`. Advanced research and diagnostics require explicit signed grants (`argus:tools:advanced` / `argus:tools:diagnostics`) plus the tool’s data scopes. Source-use checks, personal scopes and current owner consent remain separate. Do not retry a permission denial unchanged.

## Personal data boundary (registry v20)

Follow `docs/personal-examples.md` and `docs/interfaces/client-support.md` for
tested paths, exact private request templates, domain/background/export authority
and actual revocation granularity. A published Skill does not certify an untested
hosted connector. Do not share private records or download references between
participants unless the owner grant and current source rights permit it.

An Agent participant may use explicitly authorized personal domain tools under
current owner consent. It receives the per-tool `personal-result-v1` schema;
financial children retain evidence, time, source rights and audit linkage. Use
keys, versions and bound confirmations as declared, and keep user text labeled
untrusted private data. Do not share private content across participants without
the owner's authorized scope. Account revocation, deletion and identity changes
require the trusted owner flow; neither orchestration nor a confirmation flag
can perform them. See `docs/contracts/personal-tools.md`.
