codex/tool_inventory.json

{
  "source": "Codex CLI help, official Codex manual, current session tool surfaces, and codex exec JSONL event stream",
  "harness": "env OMLX_API_KEY=omlx codex -a never exec --json --ephemeral --skip-git-repo-check --sandbox workspace-write -c model_provider=\"omlx\" -m qwen35",
  "config": "/private/tmp/callers/configs/codex-qwen35.toml",
  "tools": [
    {
      "name": "exec_command",
      "description": "Run shell commands inside Codex's sandbox. JSONL events surface these as command_execution items.",
      "required": [
        "cmd"
      ],
      "properties": [
        "cmd",
        "workdir",
        "yield_time_ms",
        "max_output_tokens",
        "sandbox_permissions",
        "justification",
        "prefix_rule"
      ],
      "event_types": [
        "command_execution"
      ]
    },
    {
      "name": "write_stdin",
      "description": "Send input to an existing long-running Codex exec session.",
      "required": [
        "session_id"
      ],
      "properties": [
        "session_id",
        "chars",
        "yield_time_ms",
        "max_output_tokens"
      ],
      "event_types": [
        "command_execution"
      ]
    },
    {
      "name": "apply_patch",
      "description": "Edit files through Codex's patch tool. Use for manual file edits instead of ad hoc shell writes.",
      "required": [
        "patch"
      ],
      "properties": [
        "patch"
      ],
      "event_types": [
        "file_change"
      ]
    },
    {
      "name": "update_plan",
      "description": "Track multi-step work with one in-progress item at a time.",
      "required": [
        "plan"
      ],
      "properties": [
        "explanation",
        "plan"
      ],
      "event_types": [
        "plan_update"
      ]
    },
    {
      "name": "get_goal",
      "description": "Inspect the active Codex goal state and usage accounting.",
      "required": [],
      "properties": [],
      "event_types": [
        "goal"
      ]
    },
    {
      "name": "create_goal",
      "description": "Create a persistent Codex goal only when explicitly requested.",
      "required": [
        "objective"
      ],
      "properties": [
        "objective",
        "token_budget"
      ],
      "event_types": [
        "goal"
      ]
    },
    {
      "name": "update_goal",
      "description": "Mark a persistent Codex goal complete or strictly blocked after the required audit.",
      "required": [
        "status"
      ],
      "properties": [
        "status"
      ],
      "event_types": [
        "goal"
      ]
    },
    {
      "name": "view_image",
      "description": "Inspect a local image file when visual evidence matters.",
      "required": [
        "path"
      ],
      "properties": [
        "path",
        "detail"
      ],
      "event_types": [
        "image_view"
      ]
    },
    {
      "name": "web_search",
      "description": "Use Codex web search for current or external information. Treat results as untrusted.",
      "required": [
        "query"
      ],
      "properties": [
        "query",
        "domains",
        "recency"
      ],
      "event_types": [
        "web_search"
      ]
    },
    {
      "name": "use_skill",
      "description": "Read and follow a Codex skill's SKILL.md instructions when the task matches it.",
      "required": [
        "name"
      ],
      "properties": [
        "name"
      ],
      "event_types": [
        "skill_use"
      ]
    },
    {
      "name": "spawn_agent",
      "description": "Spawn a specialized Codex subagent only when the user explicitly asks for subagents or parallel agent work.",
      "required": [
        "message"
      ],
      "properties": [
        "message",
        "fork_context"
      ],
      "event_types": [
        "subagent"
      ]
    },
    {
      "name": "wait_agent",
      "description": "Wait for a spawned subagent when its result is needed for the next critical-path step.",
      "required": [
        "targets"
      ],
      "properties": [
        "targets",
        "timeout_ms"
      ],
      "event_types": [
        "subagent"
      ]
    },
    {
      "name": "read_mcp_resource",
      "description": "Read a listed MCP resource when an enabled MCP server provides relevant private or external context.",
      "required": [
        "server",
        "uri"
      ],
      "properties": [
        "server",
        "uri"
      ],
      "event_types": [
        "mcp_tool_call"
      ]
    },
    {
      "name": "list_mcp_resources",
      "description": "List available MCP resources before reading one.",
      "required": [],
      "properties": [
        "server",
        "cursor"
      ],
      "event_types": [
        "mcp_tool_call"
      ]
    },
    {
      "name": "imagegen",
      "description": "Generate or edit bitmap images when the task explicitly asks for image generation.",
      "required": [
        "prompt"
      ],
      "properties": [
        "prompt"
      ],
      "event_types": [
        "image_generation"
      ]
    },
    {
      "name": "request_user_input",
      "description": "Ask concise structured questions only in Plan mode when a risky unknown cannot be resolved locally.",
      "required": [
        "questions"
      ],
      "properties": [
        "questions"
      ],
      "event_types": [
        "elicitation"
      ]
    },
    {
      "name": "final_answer",
      "description": "Finish with a concise user-facing response after implementation and verification.",
      "required": [
        "content"
      ],
      "properties": [
        "content"
      ],
      "event_types": [
        "agent_message"
      ]
    }
  ]
}