Node-based terminal manager for AI coding agents — tmux-backed terminals and parallel agent sessions as draggable nodes on an infinite pan/zoom canvas. macOS, Linux, and a browser Server Edition.
TypeScript adhdagent-orchestrationai-agentscanvasclaude-codecodexcoding-agentsdeveloper-toolselectrongeminilinuxmacosmultiplexeropencodeparallel-agentsterminalterminal-emulatorterminal-multiplexertmuxworkspace-manager
import { describe, expect, it } from 'vitest'
import { TOOL_BODY_CAP, TOOL_BODY_TRUNCATED, capText, toolBody } from './chat-tool-body'
// Input shapes measured on this server's transcripts (2026-09-26):
// ExitPlanMode → {"plan": "<full markdown plan>"}
// AskUserQuestion → {"questions": [{"question", "header", "multiSelect", "options": [{"label", "description"?}]}]}
describe('toolBody — ExitPlanMode', () => {
it('returns the plan markdown verbatim', () => {
const plan = '# Fix the chat view\n\n1. Parse the plan\n2. Render it'
expect(toolBody('ExitPlanMode', { plan })).toBe(plan)
})
it('degrades to no body for a missing, non-string or blank plan', () => {
expect(toolBody('ExitPlanMode', {})).toBeUndefined()
expect(toolBody('ExitPlanMode', { plan: 42 })).toBeUndefined()
expect(toolBody('ExitPlanMode', { plan: ['a'] })).toBeUndefined()
expect(toolBody('ExitPlanMode', { plan: ' \n ' })).toBeUndefined()
expect(toolBody('ExitPlanMode', null)).toBeUndefined()
expect(toolBody('ExitPlanMode', 'plan')).toBeUndefined()
})
it('caps an oversized plan with a truncation marker', () => {
const body = toolBody('ExitPlanMode', { plan: 'x'.repeat(TOOL_BODY_CAP + 500) })!
expect(body.startsWith('x'.repeat(100))).toBe(true)
expect(body.endsWith(TOOL_BODY_TRUNCATED)).toBe(true)
expect(body.length).toBe(TOOL_BODY_CAP + TOOL_BODY_TRUNCATED.length)
})
it('never splits a surrogate pair at the cap', () => {
const plan = 'a'.repeat(TOOL_BODY_CAP - 1) + '😀' + 'tail'
const body = toolBody('ExitPlanMode', { plan })!
const kept = body.slice(0, body.length - TOOL_BODY_TRUNCATED.length)
expect(kept).toBe('a'.repeat(TOOL_BODY_CAP - 1))
})
it('leaves a plan exactly at the cap untouched', () => {
const plan = 'y'.repeat(TOOL_BODY_CAP)
expect(toolBody('ExitPlanMode', { plan })).toBe(plan)
})
})
// A side of an Edit/Write diff is file content, not markdown: the cap keeps it as the file had it.
describe('capText', () => {
it('leaves a text at the cap untouched', () => {
const s = 'z'.repeat(TOOL_BODY_CAP)
expect(capText(s)).toBe(s)
})
it('cuts an over-cap text and appends the marker — no fence a file never had', () => {
// A balanced markdown file whose cut lands inside a code block: `capBody` would close the fence.
const s = '# Doc\n```ts\n' + 'x'.repeat(TOOL_BODY_CAP) + '\n```\n'
expect(capText(s)).toBe(s.slice(0, TOOL_BODY_CAP) + TOOL_BODY_TRUNCATED)
})
it('never splits a surrogate pair at the cap', () => {
const s = 'a'.repeat(TOOL_BODY_CAP - 1) + '😀' + 'tail'
expect(capText(s)).toBe('a'.repeat(TOOL_BODY_CAP - 1) + TOOL_BODY_TRUNCATED)
})
})
describe('toolBody — AskUserQuestion', () => {
it('renders header, question and options with descriptions', () => {
const input = {
questions: [
{
question: 'Which database should we use?',
header: 'Database',
multiSelect: false,
options: [
{ label: 'Postgres', description: 'Relational, battle-tested' },
{ label: 'SQLite' }
]
}
]
}
expect(toolBody('AskUserQuestion', input)).toBe(
'**Database**\n\nWhich database should we use?\n\n- **Postgres** — Relational, battle-tested\n- **SQLite**'
)
})
it('notes a multi-select and separates several questions', () => {
const input = {
questions: [
{ question: 'Pick features', header: 'Scope', multiSelect: true, options: [{ label: 'A' }, { label: 'B' }] },
{ question: 'Ship today?', options: [{ label: 'Yes' }] }
]
}
expect(toolBody('AskUserQuestion', input)).toBe(
'**Scope**\n\nPick features _(select all that apply)_\n\n- **A**\n- **B**\n\nShip today?\n\n- **Yes**'
)
})
it('skips malformed questions and options, and degrades to no body when nothing is left', () => {
expect(toolBody('AskUserQuestion', {})).toBeUndefined()
expect(toolBody('AskUserQuestion', { questions: 'x' })).toBeUndefined()
expect(toolBody('AskUserQuestion', { questions: [] })).toBeUndefined()
expect(toolBody('AskUserQuestion', { questions: [null, 3, { header: 'H' }, { question: 7 }] })).toBeUndefined()
const input = {
questions: [
null,
{ question: 'Q?', header: 5, multiSelect: 'yes', options: [null, { label: 3 }, { label: 'ok', description: 9 }] }
]
}
expect(toolBody('AskUserQuestion', input)).toBe('Q?\n\n- **ok**')
})
it('renders a question with no options as just the question', () => {
expect(toolBody('AskUserQuestion', { questions: [{ question: 'Anything else?' }] })).toBe('Anything else?')
})
})
describe('toolBody — every other tool', () => {
it('has no body (it stays a collapsed chip)', () => {
expect(toolBody('Bash', { command: 'ls', plan: 'not a plan' })).toBeUndefined()
expect(toolBody('ToolSearch', { query: 'x' })).toBeUndefined()
})
})
describe('toolBody — hardening', () => {
it('collapses newlines inside labels and descriptions so they cannot inject list items or headings', () => {
const input = {
questions: [
{
question: 'Pick one\n\nThe question keeps its own paragraphs.',
options: [{ label: 'A\n# heading', description: 'first line\n- injected item\r\nmore' }]
}
]
}
expect(toolBody('AskUserQuestion', input)).toBe(
'Pick one\n\nThe question keeps its own paragraphs.\n\n- **A # heading** — first line - injected item more'
)
})
it('escapes emphasis characters inside the bold header and label', () => {
const input = {
questions: [{ header: 'use **all**_x', question: 'Q?', options: [{ label: 'snake_case *ptr', description: 'keeps *its* _style_' }] }]
}
expect(toolBody('AskUserQuestion', input)).toBe(
'**use \\*\\*all\\*\\*\\_x**\n\nQ?\n\n- **snake\\_case \\*ptr** — keeps *its* _style_'
)
})
it('closes an open code fence before the truncation marker', () => {
const plan = 'intro\n```ts\n' + 'x'.repeat(TOOL_BODY_CAP) + '\n```\nafter'
const body = toolBody('ExitPlanMode', { plan })!
expect(body.endsWith('\n```' + TOOL_BODY_TRUNCATED)).toBe(true)
const fences = body.split('\n').filter((l) => l.trimStart().startsWith('```')).length
expect(fences % 2).toBe(0)
})
it('does not add a fence when the kept slice has balanced fences', () => {
const plan = '```\ncode\n```\n' + 'y'.repeat(TOOL_BODY_CAP)
const body = toolBody('ExitPlanMode', { plan })!
expect(body.endsWith('y' + TOOL_BODY_TRUNCATED)).toBe(true)
})
})
// The TodoWrite input as the tool's schema states it: `{todos: [{content, status, activeForm?}]}`,
// status one of pending / in_progress / completed. No TodoWrite record exists on this server to
// measure (its CLIs, 2.1.209–2.1.316, write TaskCreate / TaskUpdate instead), so anything else
// degrades to no body — the plain chip.
describe('toolBody — TodoWrite (rich only)', () => {
const input = {
todos: [
{ content: 'Read the parser', status: 'completed', activeForm: 'Reading the parser' },
{ content: 'Write the card', status: 'in_progress', activeForm: 'Writing the card' },
{ content: 'Ship it', status: 'pending', activeForm: 'Shipping it' }
]
}
it('renders one checklist line per todo, the one in progress in bold', () => {
expect(toolBody('TodoWrite', input, { rich: true })).toBe(
'- [x] Read the parser\n- [ ] **Write the card** _(in progress)_\n- [ ] Ship it'
)
})
it('has no body without rich (the golden fixtures and the phone stay the chip)', () => {
expect(toolBody('TodoWrite', input)).toBeUndefined()
expect(toolBody('TodoWrite', input, {})).toBeUndefined()
expect(toolBody('TodoWrite', input, { rich: false })).toBeUndefined()
})
it('degrades to no body for missing or non-array todos, and skips malformed items', () => {
expect(toolBody('TodoWrite', {}, { rich: true })).toBeUndefined()
expect(toolBody('TodoWrite', { todos: 'x' }, { rich: true })).toBeUndefined()
expect(toolBody('TodoWrite', { todos: {} }, { rich: true })).toBeUndefined()
expect(toolBody('TodoWrite', { todos: [] }, { rich: true })).toBeUndefined()
expect(toolBody('TodoWrite', { todos: [null, 3, { status: 'pending' }, { content: ' ' }] }, { rich: true })).toBeUndefined()
expect(toolBody('TodoWrite', { todos: [null, { content: 'Only me', status: 'blocked' }] }, { rich: true })).toBe(
'- [ ] Only me'
)
})
it('keeps each todo on its own line and escapes emphasis inside it', () => {
const tricky = { todos: [{ content: 'fix *ptr\n- [x] injected', status: 'in_progress' }, { content: 'a_b', status: 'completed' }] }
expect(toolBody('TodoWrite', tricky, { rich: true })).toBe(
'- [ ] **fix \\*ptr - [x] injected** _(in progress)_\n- [x] a\\_b'
)
})
it('caps an oversized list', () => {
const big = { todos: Array.from({ length: 4000 }, (_, k) => ({ content: `task ${k} ${'x'.repeat(20)}`, status: 'pending' })) }
const body = toolBody('TodoWrite', big, { rich: true })!
expect(body.endsWith(TOOL_BODY_TRUNCATED)).toBe(true)
expect(body.length).toBe(TOOL_BODY_CAP + TOOL_BODY_TRUNCATED.length)
})
it('rich changes nothing for the other tools', () => {
expect(toolBody('ExitPlanMode', { plan: 'p' }, { rich: true })).toBe('p')
expect(toolBody('Bash', { command: 'ls' }, { rich: true })).toBeUndefined()
})
})