Restructure Console docs into agents, data, and workflows

[ogael-amplify]

Summary

Restructures the Amplify Console docs tab around three layers that build on each other:

1. Agents & detections — the primitives you author (agents, skills, detections)
2. Data & connections — what agents can see (projects, connections, vendor data, findings)
3. Workflows — automating it, as the culmination of the first two

A new Quickstart walks all three in one pass: connect a repo → ask an agent → write an agent → automate it.

24 pages in the Console tab, 17 new or rewritten. The legacy tab is untouched.

Structure

Group	Pages
Get Started	introduction (rewritten), quickstart (new), install-console
Agents & detections (agents/)	overview, how-agents-work, writing-an-agent, library, skills, detections, tool-reference
Data & connections (data/)	overview, projects, connections, what-agents-can-read, findings
Working interactively (interactive/)	chat, cli
Workflows (workflows/)	overview, create-a-workflow, agent-chain, triggers, outputs, running, artifacts

The two load-bearing reference pages are agents/writing-an-agent (the full AGENT.md YAML frontmatter schema, with description called out as functional rather than a comment — the workflow orchestrator reads it to brief each step) and agents/tool-reference (every tool an agent can call, which is the real ceiling on what any agent can do).

Notable moves

• The agent catalog moved out of the Workflows section to agents/library — which agents exist is a property of the platform, not of workflows.
• workflows/agents → workflows/agent-chain, now focused purely on execution semantics.
• The chat deep-dive and web-vs-CLI comparison moved out of introduction into the new Working interactively group, so the introduction can lead with the thesis.

All old paths have redirects entries in docs.json.

Capability boundaries are documented, not omitted

Several places where the product's story outruns what ships today are called out explicitly in <Warning>/<Note> callouts rather than left ambiguous:

• Vendor connections are readable only as vulnerability findings. The connector catalog spans 8 categories including ITSM and GRC, so Jira can be connected — but no agent tool can read Jira issues, CSPM config, IDP data, or GRC controls. data/what-agents-can-read carries an explicit "cannot" list.
• web_fetch is read-only (no request method or body), so there is no path for an agent to write to a third-party API. Outbound writes go through workflow outputs.
• patch is the only artifact kind produced today, though the artifact model is generic.
• Skills have no web editor — they're authored in the CLI or via the API.
• CodeQL detections are stored but not executed (only OpenGrep and policy have runtimes).
• Scheduled triggers don't exist yet.

The reasoning: docs that promise Jira reads generate support tickets we can't close. data/what-agents-can-read instead leads with what that surface genuinely does better than anything else in the stack — reachability triage of an existing scanner backlog.

Also

Replaces README.md, which was unmodified Mintlify starter-kit boilerplate pointing at the old mintlify CLI and mint.json instead of mint and docs.json — following its troubleshooting steps sent you in circles.

Verification

Rendered locally with mint dev:

• docs.json parses; all 24 Console-tab pages return HTTP 200 with correct headings
• 87 internal links and heading anchors resolve — 0 broken
• Every page has title + description frontmatter; no orphaned pages
• All 18 redirects resolve to the intended pages
• Legacy tab intact

Note for reviewers

install-console.mdx has unrelated uncommitted lab → stg URL changes in the working tree that predate this work. They are deliberately not included here — they belong in their own commit.

🤖 Generated with Claude Code

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
amplifysecurity	🟢 Ready	View Preview	Aug 3, 2026, 6:55 PM

💡 Tip: Enable Workflows to automatically generate PRs for you.