Show agent readiness and capabilities before delegation

[BradGroux]

Closes #2931.

What this fixes

Buzz currently makes an owner infer whether an agent is safe and ready to delegate to by cross-referencing the agent editor, managed-runtime state, presence, and ACP observer logs. Missing evidence can also look like unsupported behavior, while requested and effective permission modes are not visible together.

That makes the most important pre-delegation questions unnecessarily hard to answer: is the current process healthy, which runtime and model are active, which features and tools were actually reported, how risky are those tools, and is the evidence current?

What changes

The owner-only Runtime tab now includes a readiness and capability manifest built from four existing evidence sources:

• the ACP runtime catalog for installation, authentication, adapter, and static runtime facts;
• managed lifecycle state for the current process and community connection;
• current presence;
• the encrypted owner observer stream for ACP initialize, session configuration, commands, tools, model, protocol, and permission evidence.

The card shows:

• installation, authentication, process, community, presence, and observer readiness;
• runtime, version, ACP protocol, model, and provider;
• prompt and output features with explicit reported, unavailable, or unknown states;
• runtime commands, MCP source names, tool descriptors, and tool risk classes;
• requested and effective permission modes, including divergence;
• the last verification time, freshness, and known runtime limitations.

The reducer deliberately treats absent or malformed evidence as unknown. A stopped process, failed lifecycle, offline presence, closed observer, or initialize event from an older process cannot produce a fresh ready state.

Security and compatibility

The new harness observation projects only MCP server names and permission semantics. It does not forward MCP commands, arguments, environment values, credentials, executable paths, raw lifecycle errors, or config values.

Desktop parsing is strict and bounded. Older Desktop backends that do not report the new static fields remain compatible and render those facts as unknown. A non-owner E2E regression test verifies that neither the Runtime tab nor the manifest is exposed.

Verification

• cargo test -p buzz-acp — 601 unit tests and 9 lifecycle integration tests passed.
• pnpm --dir desktop test — 3,528 tests passed.
• just desktop-tauri-test — 1,638 tests passed, 14 keychain tests ignored as designed, plus 3 mixer diagnostics passed.
• Focused manifest reducer/render tests — 13 passed.
• Focused owner and non-owner Playwright smoke coverage — 2 passed.
• Workspace Rust formatting and Clippy passed.
• Tauri formatting, Clippy, and the extracted IPC projection test passed.
• Desktop typecheck, Biome across 1,638 files, text-policy checks, production build, and E2E build passed.
• Light and dark rendered output was inspected; commit-pinned screenshots are posted on this PR.

just ci reaches the Desktop file-size gate after workspace Clippy and Biome, then stops on the unchanged upstream desktop/src-tauri/src/managed_agents/runtime.rs: 2,220 lines against its existing 2,216-line ratchet. This branch does not modify that file. The branch-owned discovery test file is below its ratchet.

Non-goals

• No new connector catalog or remote MCP inventory.
• No automatic delegation or permission-policy change.
• No public capability manifest.
• No storage of secrets, full runtime configuration, or raw observer payloads.

[BradGroux]

Screenshots

Readiness and runtime evidence

The owner can see the six readiness gates alongside runtime, ACP protocol, model, and provider evidence.

Capabilities, permissions, and tools

Reported, unavailable, and unknown states remain distinct. Requested and effective permission modes are shown together, followed by commands, MCP source names, and tool risk classes.

Complete manifest

The full card fits in the existing Runtime tab without exposing the underlying observer payload.