Problem

Claude CLI runtime (src/middleware/runtimes/claude.ts) currently only handles text prompts. It needs to support inbound image attachments via the --input-format stream-json stdin protocol.

Current state

• Uses --print flag for prompt delivery — no media support
• buildEnv() returns {}
• Output parsing via --output-format stream-json --verbose --include-partial-messages

Implementation

Inbound (images only)

• Switch from --print to --input-format stream-json stdin mode when params.media is present
• Construct tool_result image content blocks:

  { "type": "image", "source": { "type": "base64", "media_type": "image/jpeg", "data": "..." } }

• Only images are supported — audio/video must be handled by middleware STT/fallback
• Declare capability: acceptsInbound: ["image/"]

Outbound

• No native media emission capability
• emitsOutbound: false

Capability declaration

readonly mediaCapabilities = {
  acceptsInbound: ["image/"],
  emitsOutbound: false,
};

Tests

Unit tests (src/middleware/runtimes/claude.test.ts)

• execute() with media: [{ mimeType: "image/jpeg", base64: "..." }] constructs correct stdin JSON
• execute() with unsupported media type (audio/video) ignores attachment
• execute() without media uses existing --print path (backwards compat)
• mediaCapabilities reports correct values

Live smoke tests (src/middleware/__smoke__/claude.live.test.ts)

• Send image attachment, verify Claude describes the image content
• Send text-only prompt, verify existing behavior unchanged

Depends on

• feat(middleware): add multimodal media attachments to AgentRuntime contract #385 — AgentRuntime multimodal contract

Related

• feat(runtimes): implement multimodal I/O for each CLI runtime #386 — per-runtime multimodal (parent tracking issue)
• research(auth+media): auth profiles and media understanding architecture #376 — auth/media restructuring