{
  "id": "capability:vision-input",
  "_kind": "Capability",
  "_file": "capabilities/capabilities/vision-input.yaml",
  "_cluster": "capabilities",
  "attributes": {
    "displayName": "Vision input",
    "description": "Ability to accept image inputs (screenshots, photos, pasted clipboard\nimages) as part of a user/agent turn. Required by interaction primitives\nthat paste or attach visual content. Distinct from `text-streaming` or\nother text-only input modalities.\n",
    "appliesToNodeKinds": [
      "ModelVersion",
      "AgentRuntimeImpl"
    ],
    "category": "modality"
  },
  "outgoingEdges": [],
  "incomingEdges": [
    {
      "from": "agent-core-impl:pikiclaw.core@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=1.0.0",
        "level": "full",
        "notes": "Camera and photo library input on mobile devices."
      }
    },
    {
      "from": "interaction-primitive:paste-image",
      "to": "capability:vision-input",
      "kind": "requires_capability"
    },
    {
      "from": "agent-version:pikiclaw@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=1.0.0",
        "level": "full",
        "notes": "Mobile camera and photo library as vision input sources."
      }
    },
    {
      "from": "agentVersion:claude:ge-0-0-0",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=2.1.156",
        "level": "full",
        "notes": "Image input is backed by official Claude multimodal documentation, with lower confidence than the Claude Code-specific docs used elsewhere."
      }
    },
    {
      "from": "agentVersion:copilot:ge-1-0-55-3",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=1.0.55-3",
        "level": "full",
        "notes": "Public image-input evidence is strongest in the Copilot SDK rather than CLI docs."
      }
    },
    {
      "from": "agentVersion:cursor:ge-0-0-0",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=0.0.0",
        "level": "full",
        "notes": "Cursor's agent overview documents image-capable reads and image generation."
      }
    },
    {
      "from": "agentVersion:gemini:ge-0-43-0",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=0.43.0",
        "level": "full",
        "notes": "Public Gemini CLI evidence describes multimodal CLI support."
      }
    },
    {
      "from": "agentVersion:omp:ge-15-5-9",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=15.5.11",
        "level": "full",
        "notes": "Public README advertises image support; technical detail is thinner."
      }
    },
    {
      "from": "agentVersion:openclaw:ge-2026-5-27",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=2026.5.27",
        "level": "full",
        "notes": "OpenClaw fallback metadata."
      }
    },
    {
      "from": "agentVersion:opencode:ge-0-0-0",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=1.15.13",
        "level": "full",
        "notes": "Public OpenCode docs lack strong technical detail for image input."
      }
    },
    {
      "from": "agentVersion:pi:ge-0-75-5",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=0.77.0",
        "level": "full",
        "notes": "Pi fallback harness."
      }
    },
    {
      "from": "agent-version:vibe-kanban@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=0.1.0",
        "level": "partial",
        "notes": "PR review screenshots and visual diffs available in the review column for agents that support vision."
      }
    },
    {
      "from": "model:amazon-nova-2-lite@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=2.0.0 <3.0.0",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:amazon-nova-2-lite-model-card"
        ]
      }
    },
    {
      "from": "model:amazon-nova-2-omni@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "level": "full"
      }
    },
    {
      "from": "model:amazon-nova-2-pro-preview@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=2.0.0 <3.0.0",
        "level": "partial",
        "evidenceSourceIds": [
          "evidence:amazon-nova-2-announcement-issue-535"
        ]
      }
    },
    {
      "from": "model:claude-fable-5@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=5.0.0 <5.1.0",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:anthropic-fable-mythos-5-models-overview-issue-991-992"
        ]
      }
    },
    {
      "from": "model:claude-mythos-5@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=5.0.0 <5.1.0",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:anthropic-fable-mythos-5-models-overview-issue-991-992"
        ]
      }
    },
    {
      "from": "model:claude-opus-4-7@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=4.7.0 <5.0.0",
        "level": "full",
        "notes": "Opus 4.7 accepts image modality inputs per its modalities list. ",
        "evidenceSourceIds": [
          "evidence:anthropic-models-doc"
        ]
      }
    },
    {
      "from": "model:cohere-embed-v4-0@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=4.0.0 <5.0.0",
        "level": "full"
      }
    },
    {
      "from": "model:command-a-plus-05-2026@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=2026.05.0 <2026.06.0",
        "level": "full"
      }
    },
    {
      "from": "model:command-a-vision@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "level": "full"
      }
    },
    {
      "from": "model:gemini-2-0-flash-preview-image-generation@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=2.0.0 <2.1.0",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:google-gemini-deprecations-issue-534"
        ]
      }
    },
    {
      "from": "model:gemini-2-5-flash-image@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=2.5.0 <2.6.0",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:google-gemini-image-models-issue-534"
        ]
      }
    },
    {
      "from": "model:gemini-2-5-flash-lite@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=2.5.0 <2.6.0",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:google-gemini-image-models-issue-534"
        ]
      }
    },
    {
      "from": "model:gemini-3-1-flash-image@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=3.1.0 <3.2.0",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:google-gemini-image-models-issue-534"
        ]
      }
    },
    {
      "from": "model:gemini-3-pro-image@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=3.0.0 <3.2.0",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:google-gemini-image-models-issue-534"
        ]
      }
    },
    {
      "from": "model:gpt-4o@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": "*",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:openai-responses-api"
        ]
      }
    },
    {
      "from": "model:gpt-5@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=5.0.0 <6.0.0",
        "level": "full",
        "notes": "GPT-5 accepts image modality inputs.",
        "evidenceSourceIds": [
          "evidence:openai-responses-api"
        ]
      }
    },
    {
      "from": "model:grok-4-3@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=4.3.0 <4.4.0",
        "level": "full"
      }
    },
    {
      "from": "model:mistral-large-3@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=3.0.0 <4.0.0",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:mistral-large-3-model-card"
        ]
      }
    },
    {
      "from": "model:mistral-medium-3-5@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=3.5.0 <4.0.0",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:mistral-medium-3-5-model-card"
        ]
      }
    },
    {
      "from": "model:mistral-small-4@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "level": "full"
      }
    },
    {
      "from": "model:claude-opus-4-8@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "versionRange": ">=4.8.0 <4.9.0",
        "level": "full",
        "evidenceSourceIds": [
          "evidence:anthropic-models-overview-claude-opus-4-8-issue-532"
        ]
      }
    },
    {
      "from": "model:pixtral-large@current",
      "to": "capability:vision-input",
      "kind": "supports",
      "attributes": {
        "level": "full"
      }
    }
  ]
}