Operations: docs and skills

Summer’s skills, templates, references and the engine class reference. 5 operations. How to call them, and what the answers mean: Operations reference.

Generated from the operation catalog: src/lib/ops/public-catalog.json in publicsummerengine at 0d85f61e2b and the summer-engine npm package 4.0.0 for engine and local operations. If something here is wrong, the operation’s definition is wrong.

Operation	What it does	Asks you first
`docs.engine_api`	Offline engine class-reference lookup — verify a property, method, signal, or constant before writing script code instead of guessing names.	Never
`docs.feedback`	Report how library entries worked out (worked, wrong, outdated, …) so Summer can fix and re-rank them; fire-and-forget, opt-out honored.	Not by default
`docs.plan`	Plan the right Summer workflow for a game-building task: skill routes, tool groups, confirmation gates, and verification steps.	Never
`docs.read`	Read one Summer skill, Studio guide or docs page by its id from docs.search, as Markdown.	Never
`docs.search`	Search Summer’s own skills, Studio guides and docs pages by describing the task in plain words.	Never

engine_api

docs.engine_api: Engine api Offline engine class-reference lookup — verify a property, method, signal, or constant before writing script code instead of guessing names.


MCP	`summer_api_docs` in the npm MCP
CLI	`summer docs engine-api`
Runs	On your computer, in the `summer-engine` npm package.
Effect	Reads only. Changes nothing.
Asks you first	No. It changes nothing.
3.x names	`summer_api_docs`

Inputs:

Input	Type	Required	Description
`class_name`	string	Yes	Engine class name, e.g. ‘BoxShape3D’, ‘CharacterBody3D’, ‘SurfaceTool’. Case-insensitive.
`member`	string	No	Optional property/method/signal/constant name to fetch just that member, e.g. ‘size’ or ‘move_and_slide’.

Input JSON schema

{
  "type": "object",
  "properties": {
    "class_name": {
      "type": "string",
      "description": "Engine class name, e.g. 'BoxShape3D', 'CharacterBody3D', 'SurfaceTool'. Case-insensitive."
    },
    "member": {
      "type": "string",
      "description": "Optional property/method/signal/constant name to fetch just that member, e.g. 'size' or 'move_and_slide'."
    }
  },
  "required": [
    "class_name"
  ],
  "additionalProperties": false
}

Example: Engine api

{
  "name": "summer_api_docs",
  "arguments": {}
}

summer docs engine-api

feedback

docs.feedback: Feedback Report how library entries worked out (worked, wrong, outdated, …) so Summer can fix and re-rank them; fire-and-forget, opt-out honored.


MCP	`summer_library_feedback` in the npm MCP
CLI	`summer docs feedback`
Runs	On your computer, in the `summer-engine` npm package.
Effect	Changes your own drafts, files or settings. Nobody else sees it yet.
Asks you first	No: Make and build is on by default in Agent permissions. When it is off, your agent asks you first.
3.x names	`summer_library_feedback`

Inputs:

Input	Type	Required	Description
`reports`	object[]	Yes	1-10 outcome reports, batched at a natural checkpoint (one call, not one per entry). Items 1 to 10.
`reports[].entry_id`	string	Yes	The library entry ID exactly as the loader printed it (kind/slug, optionally @content-hash). Never guess or reconstruct it.
`reports[].outcome`	”worked” \| “worked_with_fixes” \| “wrong” \| “outdated” \| “incomplete” \| “did_not_apply” \| “misrouted”	Yes	worked = verified in-engine (playtest/screenshot passed). worked_with_fixes = worked after your changes (say what in deviation). wrong = incorrect content. outdated = no longer matches this engine version. incomplete = missing a needed case. did_not_apply = loaded but irrelevant to the task. misrouted = the description/metadata led you here wrongly.
`reports[].note`	string	No	Optional, max 280 chars. About the ENTRY only — never the user’s project, files, or code. Length 0 to 280.
`reports[].deviation`	string	No	Optional, max 280 chars. What you did instead of / on top of the entry’s instructions. Length 0 to 280.
`engine_version`	string	Yes	The Summer Engine version in use, e.g. “4.6.1”. Length 1 to 32.
`agent_model`	string	Yes	The model you are, e.g. claude-fable-5, gpt-5.5-codex; use “unknown” if unsure. Length 1 to 64.

Input JSON schema

{
  "type": "object",
  "properties": {
    "reports": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "entry_id": {
            "type": "string",
            "pattern": "^(tool|skill|example|template|collection|reference)\\/[a-z0-9-]+(@[a-f0-9]{8,64})?$",
            "description": "The library entry ID exactly as the loader printed it (kind/slug, optionally @content-hash). Never guess or reconstruct it."
          },
          "outcome": {
            "type": "string",
            "enum": [
              "worked",
              "worked_with_fixes",
              "wrong",
              "outdated",
              "incomplete",
              "did_not_apply",
              "misrouted"
            ],
            "description": "worked = verified in-engine (playtest/screenshot passed). worked_with_fixes = worked after your changes (say what in deviation). wrong = incorrect content. outdated = no longer matches this engine version. incomplete = missing a needed case. did_not_apply = loaded but irrelevant to the task. misrouted = the description/metadata led you here wrongly."
          },
          "note": {
            "type": "string",
            "maxLength": 280,
            "description": "Optional, max 280 chars. About the ENTRY only — never the user's project, files, or code."
          },
          "deviation": {
            "type": "string",
            "maxLength": 280,
            "description": "Optional, max 280 chars. What you did instead of / on top of the entry's instructions."
          }
        },
        "required": [
          "entry_id",
          "outcome"
        ],
        "additionalProperties": false
      },
      "minItems": 1,
      "maxItems": 10,
      "description": "1-10 outcome reports, batched at a natural checkpoint (one call, not one per entry)."
    },
    "engine_version": {
      "type": "string",
      "minLength": 1,
      "maxLength": 32,
      "description": "The Summer Engine version in use, e.g. \"4.6.1\"."
    },
    "agent_model": {
      "type": "string",
      "minLength": 1,
      "maxLength": 64,
      "description": "The model you are, e.g. claude-fable-5, gpt-5.5-codex; use \"unknown\" if unsure."
    }
  },
  "required": [
    "reports",
    "engine_version",
    "agent_model"
  ],
  "additionalProperties": false
}

Example: Feedback

{
  "name": "summer_library_feedback",
  "arguments": {}
}

summer docs feedback

plan

docs.plan: Plan Plan the right Summer workflow for a game-building task: skill routes, tool groups, confirmation gates, and verification steps.


MCP	`summer_start_game_task` in the npm MCP
CLI	`summer docs plan`
Runs	On your computer, in the `summer-engine` npm package.
Effect	Reads only. Changes nothing.
Asks you first	No. It changes nothing.
3.x names	`summer_start_game_task`

Inputs:

Input	Type	Required	Description
`goal`	string	Yes	The user’s game-building goal or task.
`mode`	”auto” \| “new-game” \| “feature” \| “asset” \| “debug” \| “playtest” \| “polish” \| “ship”	No	Optional task mode override. Default `"auto"`.
`target`	”auto” \| “2d” \| “3d” \| “ui” \| “audio” \| “animation” \| “level” \| “npc” \| “multiplayer”	No	Optional content/system target override. Default `"auto"`.
`assetPolicy`	”reuse-first” \| “ask-before-paid-generation” \| “no-paid-generation” \| “generate-when-clearly-needed”	No	How aggressively to use paid asset generation. Default `"ask-before-paid-generation"`.
`verification`	”none” \| “fast” \| “full”	No	How much engine verification the agent should plan for. Default `"full"`.

Input JSON schema

{
  "type": "object",
  "properties": {
    "goal": {
      "type": "string",
      "description": "The user's game-building goal or task."
    },
    "mode": {
      "type": "string",
      "enum": [
        "auto",
        "new-game",
        "feature",
        "asset",
        "debug",
        "playtest",
        "polish",
        "ship"
      ],
      "default": "auto",
      "description": "Optional task mode override."
    },
    "target": {
      "type": "string",
      "enum": [
        "auto",
        "2d",
        "3d",
        "ui",
        "audio",
        "animation",
        "level",
        "npc",
        "multiplayer"
      ],
      "default": "auto",
      "description": "Optional content/system target override."
    },
    "assetPolicy": {
      "type": "string",
      "enum": [
        "reuse-first",
        "ask-before-paid-generation",
        "no-paid-generation",
        "generate-when-clearly-needed"
      ],
      "default": "ask-before-paid-generation",
      "description": "How aggressively to use paid asset generation."
    },
    "verification": {
      "type": "string",
      "enum": [
        "none",
        "fast",
        "full"
      ],
      "default": "full",
      "description": "How much engine verification the agent should plan for."
    }
  },
  "required": [
    "goal"
  ],
  "additionalProperties": false
}

Example: Plan

{
  "name": "summer_start_game_task",
  "arguments": {}
}

summer docs plan

read

docs.read: Read a Summer doc Read one Summer skill, Studio guide or docs page by its id from docs.search, as Markdown. Long texts are cut at maxCharacters and say so. Free.


MCP	`summer_docs_read`, a tool on the hosted MCP. Elsewhere: `summer_call` with `"op": "docs.read"`
CLI	`summer docs read`
Runs	On Summer’s servers: the hosted MCP, or the npm MCP after `summer login`. The npm MCP also runs it on your computer. Needs your Summer sign-in.
Effect	Reads only. Changes nothing.
Asks you first	No. It changes nothing.
3.x names	`summer_read_library`

Inputs:

Input	Type	Required	Description
`id`	string	Yes	skill/<slug>, guide/<id> or docs/<path>, from docs.search. Length 3 to 200.
`maxCharacters`	integer	No	Most characters of text to return. Default `24000`. Range 500 to 200000.

Input JSON schema

{
  "type": "object",
  "properties": {
    "id": {
      "description": "skill/<slug>, guide/<id> or docs/<path>, from docs.search.",
      "type": "string",
      "minLength": 3,
      "maxLength": 200,
      "pattern": "^(skill|guide|docs)\\/[A-Za-z0-9/_.-]+$"
    },
    "maxCharacters": {
      "description": "Most characters of text to return.",
      "default": 24000,
      "type": "integer",
      "minimum": 500,
      "maximum": 200000
    }
  },
  "required": [
    "id"
  ]
}

Result JSON schema

{
  "type": "object",
  "properties": {
    "id": {
      "type": "string"
    },
    "kind": {
      "type": "string",
      "enum": [
        "skill",
        "guide",
        "docs"
      ]
    },
    "title": {
      "type": "string"
    },
    "url": {
      "anyOf": [
        {
          "type": "string"
        },
        {
          "type": "null"
        }
      ]
    },
    "text": {
      "type": "string"
    },
    "truncated": {
      "type": "boolean"
    },
    "totalCharacters": {
      "type": "number"
    }
  },
  "required": [
    "id",
    "kind",
    "title",
    "url",
    "text",
    "truncated",
    "totalCharacters"
  ],
  "additionalProperties": false
}

Example: The publishing guide, start to finish.

{
  "name": "summer_docs_read",
  "arguments": {
    "id": "skill/publish-your-game"
  }
}

summer docs read --id skill/publish-your-game

search

docs.search: Search Summer docs Search Summer’s own skills, Studio guides and docs pages by describing the task in plain words. Returns ids, titles and summaries. Free. Next: docs.read with an id.


MCP	`summer_docs_search`, a tool on the hosted MCP. Elsewhere: `summer_call` with `"op": "docs.search"`
CLI	`summer docs search`
Runs	On Summer’s servers: the hosted MCP, or the npm MCP after `summer login`. The npm MCP also runs it on your computer. Needs your Summer sign-in.
Effect	Reads only. Changes nothing.
Asks you first	No. It changes nothing.
3.x names	`summer_search_library`

Inputs:

Input	Type	Required	Description
`query`	string	Yes	The task in plain words: “publish my game”, “multiplayer lobby”, “store art sizes”. Length 1 to 300.
`kind`	”skill” \| “guide” \| “docs”	No	Only skills (how to build with Summer), guides (Studio tools) or docs pages.
`limit`	integer	No	Default `8`. Range 1 to 20.

Input JSON schema

{
  "type": "object",
  "properties": {
    "query": {
      "description": "The task in plain words: \"publish my game\", \"multiplayer lobby\", \"store art sizes\".",
      "type": "string",
      "minLength": 1,
      "maxLength": 300
    },
    "kind": {
      "description": "Only skills (how to build with Summer), guides (Studio tools) or docs pages.",
      "type": "string",
      "enum": [
        "skill",
        "guide",
        "docs"
      ]
    },
    "limit": {
      "default": 8,
      "type": "integer",
      "minimum": 1,
      "maximum": 20
    }
  },
  "required": [
    "query"
  ]
}

Result JSON schema

{
  "type": "object",
  "properties": {
    "items": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "id": {
            "description": "For docs.read.",
            "type": "string"
          },
          "kind": {
            "type": "string",
            "enum": [
              "skill",
              "guide",
              "docs"
            ]
          },
          "title": {
            "type": "string"
          },
          "summary": {
            "type": "string"
          },
          "url": {
            "anyOf": [
              {
                "type": "string"
              },
              {
                "type": "null"
              }
            ]
          }
        },
        "required": [
          "id",
          "kind",
          "title",
          "summary",
          "url"
        ],
        "additionalProperties": false
      }
    }
  },
  "required": [
    "items"
  ],
  "additionalProperties": false
}

Example: Find the publishing guide.

{
  "name": "summer_docs_search",
  "arguments": {
    "query": "publish my game on Summer Games"
  }
}

summer docs search --query 'publish my game on Summer Games'

Overview

Operations

HTTP API

GDScript API

Migration and extensions

MCP

Operations: docs and skills

engine_api

feedback

plan

read

search

​engine_api

​feedback

​plan

​read

​search

engine_api

feedback

plan

read

search