Skip to main content
Glama

Dream: learn from past sessions

godot_dream

Review what Godotwright learned across sessions and propose which lessons and project hints to keep, merge, retire or rewrite. Apply only the proposals you approve.

Instructions

Look back at what Godotwright learned and did in this project, between sessions, and propose what to keep: draft lessons for Godot errors that keep coming back with no rule or lesson, project hints for API names the agent keeps guessing (use X, not Y), lessons to merge, retire or rewrite (shown often but marked wrong, or duplicates), lessons about writes that were undone again and again, rules this project marks as false alarms again and again (a suggestion only), and verdicts that contradict each other. It reads the local learning file and the project's journal, and with transcripts: true also this project's Claude Code and Codex session transcripts (only anonymous facts are kept: tool names, Godot message templates, Godot API names; no text, names, paths or code). Deterministic: no model, no network. Each proposal has a number, an id, its evidence and a reason. Nothing changes until apply (numbers or ids, or "all"): lessons go through the same scrubbing as godot_lesson_add, hints, a dated report and a short "Godotwright lessons for agents" note are kept only on this machine (the project's untracked .godot/godotwright/dreams/, or Godotwright's state folder when git tracks .godot), never in a file of the repository; the session start shows the note to the agent. godot_undo takes an apply back, lessons included. Draft lessons hold the symptom and message only: look up the cause and fix (godot_lessons, godot_api, godot_check) and rewrite them with godot_lesson_add, same title. Show the person the proposals with their evidence and apply only what they approve.

Tool call template

This tool declares no required parameters. Add optional parameters from the input schema as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {},
    "name": "godot_dream"
  }
}

Input Schema

NameRequiredDescriptionDefault
pathNoproject folder, or a file or folder inside it (relative to the server's folder; default: the current project)
applyNoproposals to apply: numbers or ids from the list, comma separated, or "all"; leave out to only list them
sinceNoonly journal events and transcripts this recent: 7d, 12h, 2w or a date (default 30d)
detailNobrief (capped lists) or full (everything); default full
transcriptsNoalso read this project's Claude Code and Codex session transcripts (default false)

Schema Changelog

Changes observed during successful MCP inspections.

  1. Addedv0.8.0

TDQS

A4.3/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations give only generic hints, and the description adds real behavioral context: deterministic, no model, no network; transcripts:"true" reads Claude Code and Codex session transcripts with only anonymous facts (tool names, message templates, API names; no text, names, paths, code); applied output is scrubbed like godot_lesson_add and written only to untracked .godot/godotwright/dreams/ or the state folder, never into a tracked repository file; the session start surfaces the note. Consistent with readOnlyHint=false and destructiveHint=false, though it does not state read cost or volume.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness3/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Purpose is front-loaded, which helps, but the text is one dense block of long run-on sentences with stacked parentheticals mixing proposal taxonomy, privacy rules, storage paths and workflow. Almost all content earns its place, yet the lack of paragraph breaks or enumeration makes it slow to parse.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a 5-parameter tool with no output schema, it covers what an agent needs: output shape (each proposal has a number, id, evidence, reason), storage locations, reversibility via godot_undo, privacy limits on transcripts, and the human-approval workflow. Nothing essential to calling it correctly appears missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so the baseline is 3; the description still adds meaning by specifying apply formats (numbers or ids from the list, comma separated, "all", omitted to only list) and by explaining that transcripts:true pulls in session transcripts with anonymous-fact filtering. since and detail semantics are largely covered by the schema, but the apply/transcripts interplay is clarified beyond it.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb and resource: looks back at what Godotwright learned and did in this project between sessions and proposes what to keep, and it enumerates the proposal categories (draft lessons for recurring errors, hints for guessed API names, merges/retirements/rewrites, undone-write lessons, false-alarm rules, contradicting verdicts). That is far more specific than sibling lesson readers/writers like godot_lessons or godot_lesson_add, so the agent can select it without opening a schema.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Describes the workflow split clearly: this tool only proposes until `apply` (numbers, ids, or "all"); draft lessons are to be looked up with godot_lessons/godot_api/godot_check and rewritten with godot_lesson_add using the same title; godot_undo reverses an apply. It also instructs showing proposals to the person and applying only approved ones. There is no explicit "do not use when" statement, but the boundaries against sibling tools are concrete.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.