@todoforai/cli 0.1.31 → 0.1.33

package/dist/todoforai-cli.js

@@ -42136,7 +42136,7 @@ var require_supports_color = __commonJS((exports, module) => {
   };
 });
 
-// node_modules/chalk/source/util.js
+// node_modules/cli-highlight/node_modules/chalk/source/util.js
 var require_util = __commonJS((exports, module) => {
   var stringReplaceAll = (string, substring, replacer) => {
     let index = string.indexOf(substring);
@@ -42175,7 +42175,7 @@ var require_util = __commonJS((exports, module) => {
   };
 });
 
-// node_modules/chalk/source/templates.js
+// node_modules/cli-highlight/node_modules/chalk/source/templates.js
 var require_templates = __commonJS((exports, module) => {
   var TEMPLATE_REGEX = /(?:\\(u(?:[a-f\d]{4}|\{[a-f\d]{1,6}\})|x[a-f\d]{2}|.))|(?:\{(~)?(\w+(?:\([^)]*\))?(?:\.\w+(?:\([^)]*\))?)*)(?:[ \t]|(?=\r?\n)))|(\})|((?:.|[\r\n\f])+?)/gi;
   var STYLE_REGEX = /(?:^|\.)(\w+)(?:\(([^)]*)\))?/g;
@@ -42287,7 +42287,7 @@ var require_templates = __commonJS((exports, module) => {
   };
 });
 
-// node_modules/chalk/source/index.js
+// node_modules/cli-highlight/node_modules/chalk/source/index.js
 var require_source = __commonJS((exports, module) => {
   var ansiStyles = require_ansi_styles();
   var { stdout: stdoutColor, stderr: stderrColor } = require_supports_color();
@@ -42744,7 +42744,14 @@ class ApiClient {
     this.apiKey = apiKey;
   }
   get headers() {
-    return { "content-type": "application/json", "x-api-key": this.apiKey };
+    const h = { "content-type": "application/json", "x-api-key": this.apiKey };
+    const extra = typeof process !== "undefined" && process.env?.TODOFORAI_EXTRA_HEADERS;
+    if (extra) {
+      try {
+        Object.assign(h, JSON.parse(extra));
+      } catch {}
+    }
+    return h;
   }
   async request(method, endpoint, body) {
     const url = `${this.apiUrl}${endpoint.replace(/^\/api\/v1/, restBasePath(this.apiKey))}`;
@@ -42857,12 +42864,27 @@ class ApiClient {
       payload.todoId = todoId;
     if (scheduledTimestamp)
       payload.scheduledTimestamp = scheduledTimestamp;
-    if (!todoId && groupTag)
+    if (groupTag)
       payload.groupTag = groupTag;
-    if (!todoId && groupName)
+    if (groupName)
       payload.groupName = groupName;
     return this.request("POST", `/api/v1/projects/${projectId}/todos`, payload);
   }
+  async registerAttachment(file, filename, opts = {}) {
+    const form = new FormData;
+    form.append("file", file, filename);
+    if (opts.todoId)
+      form.append("todoId", opts.todoId);
+    const res = await fetch(`${this.apiUrl}${restBasePath(this.apiKey)}/resources/register`, {
+      method: "POST",
+      headers: { "x-api-key": this.apiKey },
+      body: form,
+      signal: AbortSignal.timeout(60000)
+    });
+    if (!res.ok)
+      throw new Error(`API POST /resources/register failed: ${res.status} ${await res.text()}`);
+    return res.json();
+  }
   async showFile(todoId, file, filename, opts = {}) {
     const form = new FormData;
     form.append("file", file, filename);
@@ -43280,7 +43302,7 @@ var TodoStatus;
   TodoStatus2["DELETED"] = "DELETED";
 })(TodoStatus ||= {});
 // ../packages/shared-fbe/src/models.ts
-var DEFAULT_MODEL = "anthropic:anthropic/claude-sonnet-5";
+var DEFAULT_MODEL = "anthropic:anthropic/claude-opus-5";
 var MODEL_ALIASES = {
   claude: DEFAULT_MODEL
 };
@@ -43336,8 +43358,8 @@ function qualifiedModelIds(listing, filter) {
 var USAGE_WINDOWS = {
   SESSION_MS: 5 * 60 * 60 * 1000,
   WEEK_MS: 7 * 24 * 60 * 60 * 1000,
-  WEEKLY_FACTOR: 0.28,
-  SESSION_FACTOR: 1 / 7
+  WEEKLY_FACTOR: 0.35,
+  SESSION_FACTOR: 1 / 5
 };
 // ../packages/shared-fbe/src/mimetypes.json
 var mimetypes_default = {
@@ -43517,7 +43539,11 @@ var API_TOOL_TYPES = new Set([
   "api_list_mcps" /* ApiListMcps */,
   "api_get_usage" /* ApiGetUsage */,
   "api_get_balance" /* ApiGetBalance */,
-  "api_get_api_schema" /* ApiGetApiSchema */
+  "api_get_api_schema" /* ApiGetApiSchema */,
+  "api_list_board" /* ApiListBoard */,
+  "api_upsert_group" /* ApiUpsertGroup */,
+  "api_create_template" /* ApiCreateTemplate */,
+  "api_recommend_template" /* ApiRecommendTemplate */
 ]);
 var FILE_TOOL_TYPES = new Set([
   "read" /* Read */,
@@ -43959,34 +43985,36 @@ var BUSINESS_ANALYZER_AGENT = {
   ownerId: "SYSTEM",
   systemMessage: `You are a business analyst and growth advisor. Beyond analyzing and updating the business context, you provide ongoing, actionable advice for growing the company's digital presence and business.
 
-You have no previous context about this business yet \u2014 everything you know must come from the current business context markdown and what you research now. Do not assume prior knowledge or earlier conversations.
+You have no previous context about this business yet. Everything you know must come from the current business context markdown and what you research now. Do not assume prior knowledge or earlier conversations.
+
+Write dense, not long. Use concrete names, numbers and URLs, with no filler or restated context. Cut any sentence that adds no information and keep scannable structure. Use commas, colons and periods only. Never use em-dashes or semicolons anywhere, including in generated templates, notes and group descriptions.
 
-This chat lives on the user's Brand page: the business context markdown you maintain is rendered live right next to this conversation, together with the brand voice answers and the company's online-presence map. The document lives at the \`todoforai:business-context\` resource; whenever you save to it, the user sees the document change immediately \u2014 refer to it as "your brand page" (never as a file), and keep it polished since it doubles as the company's profile.
+This chat lives on the user's Brand page. The business context markdown you maintain is rendered live right next to this conversation, together with the brand voice answers and the company's online-presence map. The document lives at the \`todoforai:business-context\` resource. Whenever you save to it, the user sees the document change immediately. Refer to it as "your brand page", never as a file, and keep it polished since it doubles as the company's profile.
 
 How to analyze and update the business context:
 1. Read the current business context with \`Read todoforai:business-context\`
-2. Research: fetch websites with webfetch (follow key subpages like /about, /pricing, /contact if they exist); use google_search and google_rag to find competitors and fill knowledge gaps
-3. Merge findings into the existing context \u2014 preserve valid existing information, update conflicting facts only when the new source is clearer or more specific \u2014 and save it with \`Edit todoforai:business-context\` (or \`Write todoforai:business-context\` to create/overwrite). You MUST persist your findings this way; analysis that is only spoken in chat and never written to the brand page is lost.
+2. Research the company. Fetch websites with webfetch and follow key subpages like /about, /pricing and /contact when they exist. Use google_search and google_rag to find competitors and fill knowledge gaps
+3. Merge findings into the existing context and save it with \`Edit todoforai:business-context\`, or \`Write todoforai:business-context\` to create or overwrite it. Preserve valid existing information. Update a conflicting fact only when the new source is clearer or more specific. You MUST persist your findings this way. Analysis that is only spoken in chat and never written to the brand page is lost.
 
 Cover these sections (skip any without data):
-- **Company Info** \u2014 name, founding year, location, team size, stage
-- **Product & Services** \u2014 offerings, key features, pricing model, target audience
-- **Value Proposition** \u2014 core problem solved, unique selling points
-- **Voice & Tone** \u2014 brand personality, communication style (derive from website copy)
-- **Target Market** \u2014 customer segments, industries, geographic focus, B2B vs B2C
-- **Tech Stack** \u2014 technologies, integrations, platforms
-- **Competitors** \u2014 3-5 direct competitors with one-line comparisons
-- **Online Presence** \u2014 where the company is present: search for its accounts on X/Twitter, LinkedIn, Facebook, Instagram, YouTube, Reddit, GitHub, blog/newsletter, and any industry-relevant platforms. List each found channel as \`platform \u2014 handle/URL \u2014 one-line activity note (e.g. active, stale, unclaimed)\`; also note important channels where no presence was found
+- **Company Info**: name, founding year, location, team size, stage
+- **Product & Services**: offerings, key features, pricing model, target audience
+- **Value Proposition**: core problem solved, unique selling points
+- **Voice & Tone**: brand personality, communication style (derive from website copy)
+- **Target Market**: customer segments, industries, geographic focus, B2B vs B2C
+- **Tech Stack**: technologies, integrations, platforms
+- **Competitors**: 3-5 direct competitors with one-line comparisons
+- **Online Presence**: where the company is present. Search for its accounts on X/Twitter, LinkedIn, Facebook, Instagram, YouTube, Reddit, GitHub, blog/newsletter, and any industry-relevant platforms. List each found channel as \`platform: absolute profile URL (one-line activity note such as active, stale or unclaimed)\`. Also note important channels where no presence was found. Always record the full public URL like \`https://x.com/acme\`, never a bare handle. These links are read back to learn the company's writing voice, so an unfetchable \`@acme\` is a dead end
 
 If anything important remains unclear, add an **Open Questions** section at the end.
 
-Maintain a single machine-readable **Presence** block so the user's brand page can render a live presence map. Write it as one fenced code block tagged \`growth-plan\` containing JSON (the user never sees this raw \u2014 it powers the presence hexes on the brand page).
+Maintain a single machine-readable **Presence** block so the user's brand page can render a live presence map. Write it as one fenced code block tagged \`growth-plan\` containing JSON. The user never sees this raw. It powers the presence hexes on the brand page.
 
 \`\`\`growth-plan
 {
   "presence": {
     "linkedin-networking": { "score": 45, "handle": "https://linkedin.com/company/example", "note": "Company page exists but posts infrequently" },
-    "x-twitter-presence": { "score": 60, "handle": "@example", "note": "Active, regular posts" },
+    "x-twitter-presence": { "score": 60, "handle": "https://x.com/example", "note": "Active, regular posts" },
     "newsletter": { "score": null, "note": "No newsletter signup found" }
   }
 }
@@ -43995,46 +44023,66 @@ Maintain a single machine-readable **Presence** block so the user's brand page c
 Rules for the Presence block:
 - Keys MUST be one of these exact ids: facebook-posting, x-twitter-presence, tiktok-content, instagram-content, linkedin-networking, reddit-engagement, blog-writing, seo-optimization, backlink-building, keyword-research, aso-optimization, llm-optimization, google-ads-management, meta-ads-management, lead-outreach, crm-management, newsletter, affiliate-program, partnership-outreach, analytics-monitoring. Skip channels you did not check.
 - \`score\` is an integer 0-100, or \`null\` when you checked but found no presence. Base scores on real findings, not guesses. Include every channel you actually checked.
-- Keep it consistent with the prose Online Presence section \u2014 the JSON is the structured mirror of those findings.
+- \`handle\` is the channel's ABSOLUTE public URL such as \`https://www.linkedin.com/company/acme\` or \`https://x.com/acme\`, never a bare \`@handle\`. The brand page reads these pages back to learn the company's writing voice, so a URL that can't be fetched is worthless. Omit \`handle\` entirely when you found no account.
+- Keep it consistent with the prose Online Presence section. The JSON is the structured mirror of those findings.
 
-Growth recommendations are NOT part of this block \u2014 they live on the project board as proposed todos (\u2727 rows on their group cards; see "Recommending actionable TODOs" below). After any analysis that surfaces growth opportunities, materialize your top recommendations as proposals there; the prose Next Steps section stays human-readable advice.
+Growth recommendations are NOT part of this block. They live on the project board as proposed todos, shown as \u2727 rows on their group cards (see "Recommending actionable TODOs" below). After any analysis that surfaces growth opportunities, materialize your top recommendations as proposals there. The prose Next Steps section stays human-readable advice.
 
-Document order for the saved brand page: human-readable sections first, then the **Next Steps** section, then Open Questions (if any), and finally exactly one \`growth-plan\` block as the very last element. (Your chat reply can still end with a short spoken summary; only the saved markdown must end with the machine block.)
+Document order for the saved brand page: human-readable sections first, then the **Next Steps** section, then Open Questions if any, and finally exactly one \`growth-plan\` block as the very last element. Your chat reply can still end with a short spoken summary. Only the saved markdown must end with the machine block.
 
-Guide the user one step at a time \u2014 they should always know the single next thing to do:
-- When you need input (brand voice, tone, target audience, which channel to prioritize), prefer the question tool with short multi-choice options over asking open-ended prose questions.
-- After completing any analysis or context update, always close with a short **Next Steps** section: 2-4 concrete, prioritized tips tailored to this specific business (e.g. SEO/content gaps, social channels worth activating, positioning against the competitors you found, conversion or messaging improvements). Base tips on what you actually observed, keep each to one or two sentences, and lead with the one you'd do first.
+Guide the user one step at a time. They should always know the single next thing to do:
+- When you need input on brand voice, tone, target audience or which channel to prioritize, prefer the question tool with short multi-choice options over asking open-ended prose questions.
+- After completing any analysis or context update, always close with a short **Next Steps** section: 2-4 concrete, prioritized tips tailored to this specific business, such as SEO and content gaps, social channels worth activating, positioning against the competitors you found, or conversion and messaging improvements. Base tips on what you actually observed, keep each to one or two sentences, and lead with the one you'd do first.
 
 In follow-up conversations, act as an ongoing advisor: answer growth and digital-presence questions using the business context, refresh your analysis when asked, and suggest what to tackle next.
 
-When the user asks you to recommend how to grow (e.g. "recommend growth TODOs", the brand page's guided suggestion, the board's Recommend button), read the current business context (including the presence block), research only where information is missing or stale, refresh the presence block if findings changed, then materialize your top growth actions as recommendation cards (see below). The cards ARE the growth plan \u2014 do not write a separate recommendations list into the markdown.
+When the user asks you to recommend how to grow, for example "recommend growth TODOs", the brand page's guided suggestion, or the board's Recommend button, read the current business context including the presence block, read the board with \`api_list_board\`, research only where information is missing or stale, refresh the presence block if findings changed, then materialize your top growth actions as recommendation cards (see below). The cards ARE the growth plan. Do not write a separate recommendations list into the markdown.
 
 Strategic workstreams (board groups):
-The project board is organized into group cards, and each group should be a strategic workstream for THIS company \u2014 a front the business is (or should be) pushing on, e.g. "SEO & Content", "Outbound to agencies", "Community on Reddit", "Retention & Email". Your recommendation-generation prompt includes the board's existing groups as JSON (\`slug\`, \`name\`, and \`description\` when set \u2014 the description states the workstream's strategic intent). Treat them as the company's current strategic map:
-- FIRST fit each recommendation into an existing group when it genuinely advances that workstream \u2014 pass its slug as the \`group\` parameter of \`api_recommend_template\`. Match on strategy (read the descriptions), not keywords.
-- When your analysis surfaces a strategic direction the board doesn't cover yet, introduce it: pass a new short kebab-case \`group\` slug plus a human \`groupName\` (2-4 words, strategy-level \u2014 "Partner Channel", not "Misc" or "Growth") and a one-sentence \`groupDescription\` stating the strategic bet and its target outcome for THIS company (e.g. "Win comparison searches against Zapier and n8n to capture switchers"). The group is created with the card.
-- Keep the map tight: 3-7 workstreams total is healthy. Never invent a new group when an existing one fits, never create catch-all groups, and never mirror priority or channel-lists as groups \u2014 a group is a strategic bet, its cards are the moves.
+The project board is organized into group cards. Each group should be a strategic workstream for THIS company, a front the business is or should be pushing on, such as "SEO & Content", "Outbound to agencies", "Community on Reddit" or "Retention & Email". That board IS the company's current strategic map, so ALWAYS read it with \`api_list_board\` before you propose anything. It returns every live group with its \`slug\`, \`name\` and \`description\`, the workstream's strategic intent, together with the cards already sitting on it. Pass \`includeArchived: "true"\` when you also need the retired ones, for example before restoring one. Recommendation-generation runs also hand you the group list as JSON up front, but that is only a routing hint. It carries no cards, so it can't tell you whether a move is already proposed. Propose blind and you will duplicate what is already on the board.
+- FIRST fit each recommendation into an existing group when it genuinely advances that workstream, passing its slug as the \`group\` parameter of \`api_recommend_template\`. Match on strategy by reading the descriptions, not on keywords.
+- When your analysis surfaces a strategic direction the board doesn't cover yet, introduce it. Pass a new short kebab-case \`group\` slug plus a human \`groupName\` of 2-4 words at strategy level, like "Partner Channel" rather than "Misc" or "Growth". Add a one-sentence \`groupDescription\` stating the strategic bet and its target outcome for THIS company, for example "Win comparison searches against Zapier and n8n to capture switchers". The group is created with the card.
+- Keep the map tight. 3-7 workstreams total is healthy. Never invent a new group when an existing one fits, never create catch-all groups, and never mirror priority or channel-lists as groups. A group is a strategic bet and its cards are the moves.
+- Curate the map, don't just append to it. \`api_upsert_group\` edits a group in place. Sharpen a vague or missing \`description\` so future todos route correctly, rename a group whose cards have drifted away from its name, reorder with \`order\` to put the most important bet first, and archive a workstream the business has genuinely abandoned with \`archived: "yes"\`, or \`"no"\` to bring one back. Do this when the board tells you it's needed, such as untitled \`c-xxxx\` groups, a description that no longer matches its cards, or more than 7 live groups. Say what you changed and why in your reply, and never archive a group that still has open cards without telling the user.
 
 Recommending actionable TODOs from your recommendations:
-To materialize growth recommendations, propose todos on the project board. A proposal is a real todo in PROPOSED state \u2014 it sits as a \u2727 row on its group card until the user starts (accepts) or dismisses it. You NEVER create running todos \u2014 you only propose. Two tools, strict roles:
+To materialize growth recommendations, propose todos on the project board. A proposal is a real todo in PROPOSED state. It sits as a \u2727 row on its group card until the user accepts it by starting it, or dismisses it. You NEVER create running todos. You only propose. One tool does it: \`api_recommend_template\`.
+
+Exactly ONE \`api_recommend_template\` call per proposal: it carries both what the todo is and where it lands on the board.
+
+1. WHAT IT IS. Pass \`name\` + \`body\`, written for THIS business, and optionally \`description\`, \`short\` and \`categories\`.
+   - Reuse an existing registry template instead ONLY when it already fits this business closely with no meaningful edits: pass its id as \`templateId\` and omit \`name\`/\`body\`. If it needs tailoring, write yours inline. Do not recommend a generic playbook that ignores this company's specifics.
+
+   Language: match the business. Write user-facing text such as names, bodies and notes in the language the company itself communicates in. When the company is international or unclear, use English. Only the frame stays in English. The heading lines \`Your task:\`, \`Steps:\` and \`Data needed:\` are parsed, so keep them exactly as-is. Everything else follows the business's language.
+
+   The body is read by a PERSON deciding whether to run it, so it must be understandable at a glance. The executing agent is capable and has the business context at run time, so it will elaborate and figure out the details itself. Write a minimal brief, not a playbook. Include ONLY what the agent could not figure out on its own: the business-specific direction such as company, competitor, segment, channel and the metric to move, plus any hard constraint. Cut everything else, including background, caveats, generic how-to and restated context.
+
+   Hard limits: body \u226480 words total. One sentence of persona. One-sentence task. 2-4 steps, each a short fragment of at most 8 words naming WHAT, not how. Data needed is a comma-separated list. If a sentence wouldn't change what the agent does, delete it.
 
-1. GET A TEMPLATE (api_create_template) \u2014 generate-first, reuse only on a close fit:
-   - Default: CREATE a template tailored to THIS business with the \`api_create_template\` tool. It returns the new template id.
-   - Reuse an existing registry template ONLY when it already fits this business closely with no meaningful edits \u2014 pass its id straight to \`api_recommend_template\`. If it needs tailoring, create a new one instead \u2014 do not recommend a generic playbook that ignores this company's specifics.
+   \`name\`: 3-6 words, action-first like "Win Reddit comparison threads", with no company name prefix. Give \`description\` and \`short\` one tight sentence each.
 
-   Write every generated body to be unmistakably about THIS company \u2014 bake the specifics from the business context into the task itself: the company name, its actual competitors, the target segment, the specific channel/platform, and the concrete metric to move. A reader should never mistake it for a generic template.
+   Example body at the right density (adapt specifics, keep the brevity):
+   "You are a B2B outbound specialist for dev-tool companies.
+   Your task:
+   Land Acme in 10 agency partner conversations this month.
+   Steps:
+   - Shortlist 30 EU automation agencies
+   - Draft one-line personalized openers
+   - Send, track replies in a sheet
+   Data needed:
+   Acme pricing page, current partner list, founder's LinkedIn"
 
-   Body structure (REQUIRED \u2014 the body is split into the agent's system prompt and the first user turn by section headings, so follow this exactly):
-   - Start with the persona/expertise and any Rules as plain prose/paragraphs at the top (no heading needed) \u2014 this becomes the agent's system prompt.
-   - Then the request, using these exact standalone heading lines (letters only, each on its own line, ending with a colon): \`Your task:\`, \`Steps:\`, \`Data needed:\`. Everything under these headings becomes the first user turn.
-   - Headings must match exactly (e.g. \`Your task:\`, not \`Your task (SEO):\`) or the split breaks. Put the business-specific detail in the body text under each heading, never in the heading itself.
+   Body structure (REQUIRED). The body is split into the agent's system prompt and the first user turn by section headings, so follow this exactly:
+   - Start with the persona, expertise and any Rules as plain prose at the top with no heading. This becomes the agent's system prompt.
+   - Then write the request using these exact standalone heading lines, letters only, each on its own line and ending with a colon: \`Your task:\`, \`Steps:\`, \`Data needed:\`. Everything under these headings becomes the first user turn.
+   - Headings must match exactly or the split breaks. Write \`Your task:\`, not \`Your task (SEO):\`. Put the business-specific detail in the body text under each heading, never in the heading itself.
 
-2. RECOMMEND (api_recommend_template \u2014 proposes the template as a PROPOSED todo on the project; never creates a running todo):
-   Call \`api_recommend_template\` with the template id, a one-line \`note\` (the WHY for THIS business \u2014 it becomes the proposal's tooltip), a \`priority\` of high, medium or low, and the \`group\` slug of the strategic workstream it belongs to (see "Strategic workstreams" above \u2014 reuse an existing slug when one fits; a new slug + \`groupName\` creates the workstream). Project and business context default to the current session, so pass them only when targeting a different one. The proposal appears as a \u2727 row on that group's board card for the user to review before any credits are spent. Re-proposing the same template refreshes the open proposal instead of duplicating it; dismissed proposals since your last run are fed back to you \u2014 don't re-propose those, steer elsewhere.
+2. WHERE IT LANDS, in that same call. It becomes a PROPOSED todo on the project and never a running todo:
+   Give it a one-line \`note\` of roughly 15 words at most giving the WHY for THIS business, which becomes the proposal's tooltip, a \`priority\` of high, medium or low, and the \`group\` slug of the strategic workstream it belongs to. See "Strategic workstreams" above. Reuse an existing slug when one fits. A new slug plus \`groupName\` creates the workstream. Project and business context default to the current session, so pass them only when targeting a different one. The proposal appears as a \u2727 row on that group's board card for the user to review before any credits are spent. Re-proposing the same \`templateId\` refreshes that open proposal instead of duplicating it; an inline proposal is always a new one, so check the board you just read before repeating a move that is already there. Dismissed proposals since your last run are fed back to you. Don't re-propose those, steer elsewhere.
 
-   Both tools run in-process and need no machine or shell \u2014 never shell out to \`todoregistry-cli\` / \`todoforai-cli\` for this.
+   It runs in-process and needs no machine or shell. Never shell out to \`todoregistry-cli\` or \`todoforai-cli\` for this.
 
-3. Propose 3-6 todos (highest priority first). Then reply with a one-line summary per proposal grouped by workstream, and tell the user they're waiting on the board for review.`,
+3. Propose 3-6 todos, highest priority first. Make sure a few of them are the trivial, high-probability moves for this company, the strategic directions its owner would nod at instantly, alongside any cleverer bets. Then reply with a one-line summary per proposal grouped by workstream, and tell the user they're waiting on the board for review.`,
   mcpConfigs: {},
   edgesMcpConfigs: {},
   permissions: {
@@ -44221,7 +44269,7 @@ import { parseArgs } from "util";
 // package.json
 var package_default = {
   name: "@todoforai/cli",
-  version: "0.1.31",
+  version: "0.1.33",
   type: "module",
   bin: {
     "todoforai-cli": "bin/todoforai-cli.js",
@@ -44240,6 +44288,7 @@ var package_default = {
   },
   dependencies: {
     "@todoforai/update-notifier": "^0.1.0",
+    chalk: "^6.0.0",
     "cli-highlight": "^2.1.11",
     "diff-match-patch": "^1.0.5",
     ws: "^8.18.0"
@@ -44258,35 +44307,35 @@ function getEnv(name) {
 }
 function printUsage() {
   process.stderr.write(`
-todoforai-cli \u2014 TODOforAI CLI (Bun). Aliases: tfa-cli, todoai.
+tfa-cli \u2014 TODOforAI CLI (Bun). Legacy aliases: todoforai-cli, todoai.
 
 Usage:
-  todoforai-cli login                          # Browser-based device auth
-  todoforai-cli "prompt text"                  # Prompt as argument
-  todoforai-cli -n "Quick task"               # Non-interactive (run and exit)
-  echo "content" | todoforai-cli              # Pipe from stdin
-  todoforai-cli --path /my/project "Fix bug"  # Explicit workspace path
-  todoforai-cli -c ["prompt"]                 # Resume last todo (optional prompt sent on attach)
-  todoforai-cli --resume <todo-id> ["prompt"] # Resume specific todo (optional prompt sent on attach)
-  todoforai-cli --inspect <todo-id>[@<slice>] # Read chat log. <slice> = -3:, :1, 5:10, 7  (Python-style)
-  todoforai-cli start <id>                    # Start a TODO from the registry (todoregistry.com)
-  todoforai-cli --list-agents                 # List available agents and exit
-  todoforai-cli --list-models [filter]        # List models usable with --model and exit
-  todoforai-cli agent update <agent> model=<model>    # Update agent settings (see 'agent --help')
-  todoforai-cli list [-n 30] [--cursor N] [--all] [--status S]  # List todos (paginated); see 'list --help'
-  todoforai-cli status <todo-id> <STATUS>     # Update a todo's status (run 'status --help' for the full list)
-  todoforai-cli delete <todo-id>              # Permanently delete a todo
-  todoforai-cli addmessage <todo-id> "text"  # Add a message to an existing todo
-  todoforai-cli show <file|-> [todo-id]     # Show a file in the chat (rendered by mimetype; - reads stdin)
+  tfa-cli login                          # Browser-based device auth
+  tfa-cli "prompt text"                  # Prompt as argument
+  tfa-cli -n "Quick task"               # Non-interactive (run and exit)
+  echo "content" | tfa-cli              # Pipe from stdin
+  tfa-cli --path /my/project "Fix bug"  # Explicit workspace path
+  tfa-cli -c ["prompt"]                 # Resume last todo (optional prompt sent on attach)
+  tfa-cli --resume <todo-id> ["prompt"] # Resume specific todo (optional prompt sent on attach)
+  tfa-cli --inspect <todo-id>[@<slice>] # Read chat log. <slice> = -3:, :1, 5:10, 7  (Python-style)
+  tfa-cli start <id>                    # Start a TODO from the registry (todoregistry.com)
+  tfa-cli --list-agents                 # List available agents and exit
+  tfa-cli --list-models [filter]        # List models usable with --model and exit
+  tfa-cli agent update <agent> model=<model>    # Update agent settings (see 'agent --help')
+  tfa-cli list [-n 30] [--cursor N] [--all] [--status S]  # List todos (paginated); see 'list --help'
+  tfa-cli status <todo-id> <STATUS>     # Update a todo's status (run 'status --help' for the full list)
+  tfa-cli delete <todo-id>              # Permanently delete a todo
+  tfa-cli addmessage <todo-id> "text"  # Add a message to an existing todo
+  tfa-cli show <file|-> [todo-id]     # Show a file in the chat (rendered by mimetype; - reads stdin)
                                             #   [--title T] [--alias A] [--mime M] [--card <name>] [--json]
-  todoforai-cli show list [todo-id]         # List show blocks (ref, title, mime/url, card)
+  tfa-cli show list [todo-id]         # List show blocks (ref, title, mime/url, card)
                                             #   [--project <id>] [--card <name>] [--json]
                                             #   no todo-id + --project (or $TODOFORAI_PROJECT_ID) = every todo
-  todoforai-cli open <url> [todo-id]        # Show a live http(s) url in the chat as a preview
+  tfa-cli open <url> [todo-id]        # Show a live http(s) url in the chat as a preview
                                             #   [--title T] [--alias A] [--json]
-  todoforai-cli recommend --template <id>    # Add a template as a recommendation card (see 'todoregistry-cli create')
-  todoforai-cli claim mint --seed <projectId> [--emails a@x,b@y] [--ttl <sec>]  # Mint /claim/<token> ownership links for a project you own
-  todoforai-cli next [--direction "<text>"]  # Ask the analyzer for growth recommendation cards (optional free-text steer)
+  tfa-cli recommend --template <id>    # Add a template as a recommendation card (see 'todoregistry-cli create')
+  tfa-cli claim mint --seed <projectId> [--emails a@x,b@y] [--ttl <sec>]  # Mint /claim/<token> ownership links for a project you own
+  tfa-cli next [--direction "<text>"]  # Ask the analyzer for growth recommendation cards (optional free-text steer)
 
 Options:
   --path <dir>                    Workspace path (default: cwd)
@@ -44310,6 +44359,8 @@ Options:
   --allow-all                     Set permissions to allow all tools (no approval needed)
   --raw-sysmsg <file>             Use file contents verbatim as system prompt (new TODO only)
   --no-watch                      Create todo and exit
+  --isolated                      Spawn an ephemeral, task-scoped bridge: the agent sees ONLY
+                                  this machine (workspace = --path/cwd); session dies with the CLI
   --no-bridge                     Do not auto-spawn bridge
   --no-edge                       Deprecated alias for --no-bridge
   --json                          Output as JSON
@@ -44317,6 +44368,7 @@ Options:
   --format-anthropic              'inspect --json': Anthropic-style shape (tool_result in next user msg); attachment sources are uri-typed, so not a 1:1 messages.create input
   --safe                          Validate API key upfront
   --debug, -d                     Debug output
+  --debug-dump                    Attach LLM request debug info per turn (requires server grant)
   --show-config                   Show config
   --reset-config                  Reset config file
   --version, -v                   Print version and exit
@@ -44334,7 +44386,7 @@ var STATUS_HELP = {
 };
 function printStatusHelp() {
   process.stderr.write(`
-todoforai-cli status <todo-id> <STATUS>
+tfa-cli status <todo-id> <STATUS>
 
 Common statuses:
 ${Object.entries(STATUS_HELP).map(([s, d]) => `  ${s.padEnd(18)}${d}`).join(`
@@ -44380,6 +44432,8 @@ function parseCliArgs() {
       "allow-all": { type: "boolean", default: false },
       "raw-sysmsg": { type: "string" },
       "no-watch": { type: "boolean", default: false },
+      isolated: { type: "boolean", default: false },
+      "debug-dump": { type: "boolean", default: false },
       "no-bridge": { type: "boolean", default: false },
       "no-edge": { type: "boolean", default: false },
       json: { type: "boolean", default: false },
@@ -44770,14 +44824,12 @@ function bridgeRunArgs(apiUrl) {
   const url = parseApiUrl(apiUrl);
   if (!url)
     return [];
-  if (isLocalHost(url.hostname)) {
+  if (url.hostname && url.hostname !== "api.todofor.ai") {
     const args = ["--host", url.hostname];
     if (url.port)
       args.push("--port", url.port);
     return withProfile(args, apiUrl);
   }
-  if (url.hostname && url.hostname !== "api.todofor.ai")
-    return withProfile(["--host", url.hostname], apiUrl);
   return [];
 }
 function bridgeLoginArgs(apiUrl) {
@@ -45175,7 +45227,9 @@ var DETAILED_ONLY = new Set([
   "stop_sequence",
   "permissions",
   "isPublic",
-  "metadata"
+  "metadata",
+  "systemPrompt",
+  "toolSchemas"
 ]);
 var DEBUG_ONLY = new Set([
   "runMeta",
@@ -45303,6 +45357,10 @@ function toAnthropicShape(messages, mode = "default", format = "compact") {
         msg.id = m.id;
       if (m.createdAt)
         msg.createdAt = m.createdAt;
+      if (m.systemPrompt !== undefined)
+        msg.systemPrompt = m.systemPrompt;
+      if (m.toolSchemas !== undefined)
+        msg.toolSchemas = m.toolSchemas;
     }
     if (m.role === "user") {
       const atts = (m.attachments ?? []).filter((a) => !prevToolAttachmentIds.has(a.id)).map((a) => {
@@ -45676,45 +45734,594 @@ Please choose an agent:
 // src/diff-view.ts
 var import_diff_match_patch = __toESM(require_diff_match_patch(), 1);
 var import_cli_highlight = __toESM(require_dist(), 1);
-var import_chalk = __toESM(require_source(), 1);
 import { extname } from "path";
+
+// node_modules/chalk/source/utilities.js
+function stringReplaceAll(string, substring, postfix) {
+  let index = string.indexOf(substring);
+  if (index === -1) {
+    return string;
+  }
+  const substringLength = substring.length;
+  let endIndex = 0;
+  let returnValue = "";
+  do {
+    returnValue += string.slice(endIndex, index) + substring + postfix;
+    endIndex = index + substringLength;
+    index = string.indexOf(substring, endIndex);
+  } while (index !== -1);
+  returnValue += string.slice(endIndex);
+  return returnValue;
+}
+function stringEncaseCRLFWithFirstIndex(string, prefix, postfix, index) {
+  let endIndex = 0;
+  let returnValue = "";
+  do {
+    const isGotCR = string[index - 1] === "\r";
+    returnValue += string.slice(endIndex, isGotCR ? index - 1 : index) + prefix + (isGotCR ? `\r
+` : `
+`) + postfix;
+    endIndex = index + 1;
+    index = string.indexOf(`
+`, endIndex);
+  } while (index !== -1);
+  returnValue += string.slice(endIndex);
+  return returnValue;
+}
+
+// node_modules/chalk/source/vendor/ansi-styles/index.js
+var ANSI_BACKGROUND_OFFSET = 10;
+var ANSI_UNDERLINE_OFFSET = 20;
+var wrapAnsi16 = (offset = 0) => (code) => `\x1B[${code + offset}m`;
+var wrapAnsi256 = (offset = 0) => (code) => `\x1B[${38 + offset};5;${code}m`;
+var wrapAnsi16m = (offset = 0) => (red, green, blue) => `\x1B[${38 + offset};2;${red};${green};${blue}m`;
+var wrapUnderlineAnsi = (code) => `\x1B[58;5;${code < 90 ? code - 30 : code - 90 + 8}m`;
+var styles = {
+  modifier: {
+    reset: [0, 0],
+    bold: [1, 22],
+    dim: [2, 22],
+    italic: [3, 23],
+    underline: [4, 24],
+    underlineDouble: ["4:2", 24],
+    underlineCurly: ["4:3", 24],
+    underlineDotted: ["4:4", 24],
+    underlineDashed: ["4:5", 24],
+    overline: [53, 55],
+    inverse: [7, 27],
+    hidden: [8, 28],
+    strikethrough: [9, 29]
+  },
+  color: {
+    black: [30, 39],
+    red: [31, 39],
+    green: [32, 39],
+    yellow: [33, 39],
+    blue: [34, 39],
+    magenta: [35, 39],
+    cyan: [36, 39],
+    white: [37, 39],
+    blackBright: [90, 39],
+    gray: [90, 39],
+    grey: [90, 39],
+    redBright: [91, 39],
+    greenBright: [92, 39],
+    yellowBright: [93, 39],
+    blueBright: [94, 39],
+    magentaBright: [95, 39],
+    cyanBright: [96, 39],
+    whiteBright: [97, 39]
+  },
+  bgColor: {
+    bgBlack: [40, 49],
+    bgRed: [41, 49],
+    bgGreen: [42, 49],
+    bgYellow: [43, 49],
+    bgBlue: [44, 49],
+    bgMagenta: [45, 49],
+    bgCyan: [46, 49],
+    bgWhite: [47, 49],
+    bgBlackBright: [100, 49],
+    bgGray: [100, 49],
+    bgGrey: [100, 49],
+    bgRedBright: [101, 49],
+    bgGreenBright: [102, 49],
+    bgYellowBright: [103, 49],
+    bgBlueBright: [104, 49],
+    bgMagentaBright: [105, 49],
+    bgCyanBright: [106, 49],
+    bgWhiteBright: [107, 49]
+  },
+  underlineColor: {
+    underlineBlack: ["58;5;0", 59],
+    underlineRed: ["58;5;1", 59],
+    underlineGreen: ["58;5;2", 59],
+    underlineYellow: ["58;5;3", 59],
+    underlineBlue: ["58;5;4", 59],
+    underlineMagenta: ["58;5;5", 59],
+    underlineCyan: ["58;5;6", 59],
+    underlineWhite: ["58;5;7", 59],
+    underlineBlackBright: ["58;5;8", 59],
+    underlineGray: ["58;5;8", 59],
+    underlineGrey: ["58;5;8", 59],
+    underlineRedBright: ["58;5;9", 59],
+    underlineGreenBright: ["58;5;10", 59],
+    underlineYellowBright: ["58;5;11", 59],
+    underlineBlueBright: ["58;5;12", 59],
+    underlineMagentaBright: ["58;5;13", 59],
+    underlineCyanBright: ["58;5;14", 59],
+    underlineWhiteBright: ["58;5;15", 59]
+  }
+};
+var modifierNames = Object.keys(styles.modifier);
+var foregroundColorNames = Object.keys(styles.color);
+var backgroundColorNames = Object.keys(styles.bgColor);
+var underlineColorNames = Object.keys(styles.underlineColor);
+var colorNames = [...foregroundColorNames, ...backgroundColorNames];
+function assembleStyles() {
+  const codes = new Map;
+  for (const [groupName, group] of Object.entries(styles)) {
+    for (const [styleName, style] of Object.entries(group)) {
+      styles[styleName] = {
+        open: `\x1B[${style[0]}m`,
+        close: `\x1B[${style[1]}m`
+      };
+      group[styleName] = styles[styleName];
+      codes.set(Number.parseInt(style[0], 10), style[1]);
+    }
+    Object.defineProperty(styles, groupName, {
+      value: group,
+      enumerable: false
+    });
+  }
+  Object.defineProperty(styles, "codes", {
+    value: codes,
+    enumerable: false
+  });
+  styles.color.close = "\x1B[39m";
+  styles.bgColor.close = "\x1B[49m";
+  styles.underlineColor.close = "\x1B[59m";
+  styles.color.ansi = wrapAnsi16();
+  styles.color.ansi256 = wrapAnsi256();
+  styles.color.ansi16m = wrapAnsi16m();
+  styles.bgColor.ansi = wrapAnsi16(ANSI_BACKGROUND_OFFSET);
+  styles.bgColor.ansi256 = wrapAnsi256(ANSI_BACKGROUND_OFFSET);
+  styles.bgColor.ansi16m = wrapAnsi16m(ANSI_BACKGROUND_OFFSET);
+  styles.underlineColor.ansi = wrapUnderlineAnsi;
+  styles.underlineColor.ansi256 = wrapAnsi256(ANSI_UNDERLINE_OFFSET);
+  styles.underlineColor.ansi16m = wrapAnsi16m(ANSI_UNDERLINE_OFFSET);
+  Object.defineProperties(styles, {
+    rgbToAnsi256: {
+      value(red, green, blue) {
+        if (red === green && green === blue) {
+          if (red < 8) {
+            return 16;
+          }
+          if (red > 248) {
+            return 231;
+          }
+          return Math.round((red - 8) / 247 * 24) + 232;
+        }
+        return 16 + 36 * Math.round(red / 255 * 5) + 6 * Math.round(green / 255 * 5) + Math.round(blue / 255 * 5);
+      },
+      enumerable: false
+    },
+    hexToRgb: {
+      value(hex) {
+        const matches = /[\da-f]{6}|[\da-f]{3}/i.exec(hex.toString(16));
+        if (!matches) {
+          return [0, 0, 0];
+        }
+        let [colorString] = matches;
+        if (colorString.length === 3) {
+          colorString = [...colorString].map((character) => character + character).join("");
+        }
+        const integer = Number.parseInt(colorString, 16);
+        return [
+          integer >> 16 & 255,
+          integer >> 8 & 255,
+          integer & 255
+        ];
+      },
+      enumerable: false
+    },
+    hexToAnsi256: {
+      value: (hex) => styles.rgbToAnsi256(...styles.hexToRgb(hex)),
+      enumerable: false
+    },
+    ansi256ToAnsi: {
+      value(code) {
+        if (code < 8) {
+          return 30 + code;
+        }
+        if (code < 16) {
+          return 90 + (code - 8);
+        }
+        let red;
+        let green;
+        let blue;
+        if (code >= 232) {
+          red = ((code - 232) * 10 + 8) / 255;
+          green = red;
+          blue = red;
+        } else {
+          code -= 16;
+          const remainder = code % 36;
+          red = Math.floor(code / 36) / 5;
+          green = Math.floor(remainder / 6) / 5;
+          blue = remainder % 6 / 5;
+        }
+        const value = Math.max(red, green, blue) * 2;
+        if (value === 0) {
+          return 30;
+        }
+        let result = 30 + (Math.round(blue) << 2 | Math.round(green) << 1 | Math.round(red));
+        if (value === 2) {
+          result += 60;
+        }
+        return result;
+      },
+      enumerable: false
+    },
+    rgbToAnsi: {
+      value: (red, green, blue) => styles.ansi256ToAnsi(styles.rgbToAnsi256(red, green, blue)),
+      enumerable: false
+    },
+    hexToAnsi: {
+      value: (hex) => styles.ansi256ToAnsi(styles.hexToAnsi256(hex)),
+      enumerable: false
+    }
+  });
+  return styles;
+}
+var ansiStyles = assembleStyles();
+var ansi_styles_default = ansiStyles;
+
+// node_modules/chalk/source/vendor/supports-color/index.js
+import process2 from "process";
+import os3 from "os";
+import tty from "tty";
+function hasFlag(flag, argv = globalThis.Deno ? globalThis.Deno.args : process2.argv) {
+  const prefix = flag.startsWith("-") ? "" : flag.length === 1 ? "-" : "--";
+  const position = argv.indexOf(prefix + flag);
+  const terminatorPosition = argv.indexOf("--");
+  return position !== -1 && (terminatorPosition === -1 || position < terminatorPosition);
+}
+var { env } = process2;
+var flagForceColor;
+if (hasFlag("no-color") || hasFlag("no-colors") || hasFlag("color=false") || hasFlag("color=never")) {
+  flagForceColor = 0;
+} else if (hasFlag("color") || hasFlag("colors") || hasFlag("color=true") || hasFlag("color=always")) {
+  flagForceColor = 1;
+}
+function hasNumericForceColor() {
+  return /^\d+$/.test(env.FORCE_COLOR);
+}
+function envForceColor() {
+  if (!("FORCE_COLOR" in env)) {
+    return;
+  }
+  if (env.FORCE_COLOR === "false") {
+    return 0;
+  }
+  if (env.FORCE_COLOR === "true" || env.FORCE_COLOR.length === 0) {
+    return 1;
+  }
+  if (!hasNumericForceColor()) {
+    return;
+  }
+  return Math.min(Number.parseInt(env.FORCE_COLOR, 10), 3);
+}
+function translateLevel(level) {
+  if (level === 0) {
+    return false;
+  }
+  return {
+    level,
+    hasBasic: true,
+    has256: level >= 2,
+    has16m: level >= 3
+  };
+}
+function _supportsColor(haveStream, { streamIsTTY, sniffFlags = true } = {}) {
+  const noFlagForceColor = envForceColor();
+  if (noFlagForceColor !== undefined) {
+    flagForceColor = noFlagForceColor;
+  }
+  const forceColor = sniffFlags ? flagForceColor : noFlagForceColor;
+  if (forceColor === 0) {
+    return 0;
+  }
+  if (sniffFlags) {
+    if (hasFlag("color=16m") || hasFlag("color=full") || hasFlag("color=truecolor")) {
+      return 3;
+    }
+    if (hasFlag("color=256")) {
+      return 2;
+    }
+  }
+  if (forceColor !== undefined && hasNumericForceColor()) {
+    return forceColor;
+  }
+  if ("TF_BUILD" in env && "AGENT_NAME" in env) {
+    return 1;
+  }
+  if (haveStream && !streamIsTTY && forceColor === undefined) {
+    return 0;
+  }
+  const min = forceColor || 0;
+  if (env.TERM === "dumb") {
+    return min;
+  }
+  if (process2.platform === "win32") {
+    const osRelease = os3.release().split(".");
+    if (Number(osRelease[0]) >= 10 && Number(osRelease[2]) >= 10586) {
+      return Number(osRelease[2]) >= 14931 ? 3 : 2;
+    }
+    return 1;
+  }
+  if ("CI" in env) {
+    if (["GITHUB_ACTIONS", "GITEA_ACTIONS", "CIRCLECI"].some((key) => (key in env))) {
+      return 3;
+    }
+    if (["TRAVIS", "APPVEYOR", "GITLAB_CI", "BUILDKITE", "DRONE"].some((sign) => (sign in env)) || env.CI_NAME === "codeship") {
+      return 1;
+    }
+    return min;
+  }
+  if ("TEAMCITY_VERSION" in env) {
+    return /^(?:9\.0*[1-9]\d*\.|\d{2,}\.)/.test(env.TEAMCITY_VERSION) ? 1 : 0;
+  }
+  if (env.COLORTERM === "truecolor") {
+    return 3;
+  }
+  if (env.TERM === "xterm-kitty") {
+    return 3;
+  }
+  if (env.TERM === "xterm-ghostty") {
+    return 3;
+  }
+  if (env.TERM === "wezterm") {
+    return 3;
+  }
+  if ("TERM_PROGRAM" in env) {
+    const version = Number.parseInt((env.TERM_PROGRAM_VERSION || "").split(".", 1)[0], 10);
+    switch (env.TERM_PROGRAM) {
+      case "iTerm.app": {
+        return version >= 3 ? 3 : 2;
+      }
+      case "Apple_Terminal": {
+        return 2;
+      }
+    }
+  }
+  if (/-256(?:color)?$/i.test(env.TERM)) {
+    return 2;
+  }
+  if (/^screen|^xterm|^vt100|^vt220|^rxvt|color|ansi|cygwin|linux/i.test(env.TERM)) {
+    return 1;
+  }
+  if ("COLORTERM" in env) {
+    return 1;
+  }
+  return min;
+}
+function createSupportsColor(stream, options = {}) {
+  const level = _supportsColor(stream, {
+    streamIsTTY: stream && stream.isTTY,
+    ...options
+  });
+  return translateLevel(level);
+}
+var supportsColor = {
+  stdout: createSupportsColor({ isTTY: tty.isatty(1) }),
+  stderr: createSupportsColor({ isTTY: tty.isatty(2) })
+};
+var supports_color_default = supportsColor;
+
+// node_modules/chalk/source/index.js
+var { stdout: stdoutColor, stderr: stderrColor } = supports_color_default;
+var GENERATOR = Symbol("GENERATOR");
+var STYLER = Symbol("STYLER");
+var IS_EMPTY = Symbol("IS_EMPTY");
+var LEVEL = Symbol("LEVEL");
+var styles2 = Object.create(null);
+var assertValidLevel = (level) => {
+  if (!Number.isSafeInteger(level) || level < 0 || level > 3) {
+    throw new Error("The `level` should be an integer from 0 to 3");
+  }
+};
+var levelDescriptor = {
+  enumerable: true,
+  get() {
+    return this[LEVEL];
+  },
+  set(level) {
+    assertValidLevel(level);
+    this[LEVEL] = level;
+  }
+};
+var applyOptions = (object, options = {}) => {
+  if (options.level !== undefined) {
+    assertValidLevel(options.level);
+  }
+  const colorLevel = stdoutColor ? stdoutColor.level : 0;
+  object[LEVEL] = options.level === undefined ? colorLevel : options.level;
+};
+var chalkFactory = (options) => {
+  const chalk = (...strings) => strings.join(" ");
+  applyOptions(chalk, options);
+  Object.setPrototypeOf(chalk, createChalk.prototype);
+  return chalk;
+};
+function createChalk(options) {
+  return chalkFactory(options);
+}
+Object.setPrototypeOf(createChalk.prototype, Function.prototype);
+for (const [styleName, style] of Object.entries(ansi_styles_default)) {
+  styles2[styleName] = {
+    get() {
+      const builder = createBuilder(this, createStyler(style.open, style.close, this[STYLER]), this[IS_EMPTY]);
+      Object.defineProperty(this, styleName, { value: builder });
+      return builder;
+    }
+  };
+}
+styles2.visible = {
+  get() {
+    const builder = createBuilder(this, this[STYLER], true);
+    Object.defineProperty(this, "visible", { value: builder });
+    return builder;
+  }
+};
+var createModelConverters = (model, type) => {
+  const style = ansi_styles_default[type];
+  if (model === "rgb") {
+    const ansi2 = (red, green, blue) => style.ansi(ansi_styles_default.rgbToAnsi(red, green, blue));
+    const ansi256 = (red, green, blue) => style.ansi256(ansi_styles_default.rgbToAnsi256(red, green, blue));
+    return [ansi2, ansi2, ansi256, style.ansi16m];
+  }
+  if (model === "hex") {
+    const ansi2 = (hex) => style.ansi(ansi_styles_default.hexToAnsi(hex));
+    const ansi256 = (hex) => style.ansi256(ansi_styles_default.hexToAnsi256(hex));
+    return [ansi2, ansi2, ansi256, (hex) => style.ansi16m(...ansi_styles_default.hexToRgb(hex))];
+  }
+  const ansi = (code) => style.ansi(ansi_styles_default.ansi256ToAnsi(code));
+  return [ansi, ansi, style.ansi256, style.ansi256];
+};
+var usedModels = ["rgb", "hex", "ansi256"];
+for (const model of usedModels) {
+  const capitalizedModel = model[0].toUpperCase() + model.slice(1);
+  for (const [styleName, type] of [
+    [model, "color"],
+    ["bg" + capitalizedModel, "bgColor"],
+    ["underline" + capitalizedModel, "underlineColor"]
+  ]) {
+    const { close } = ansi_styles_default[type];
+    const converters = createModelConverters(model, type);
+    styles2[styleName] = {
+      get() {
+        const styleFunction = function(first, second, third) {
+          const open = converters[this.level](first, second, third);
+          return createBuilder(this, createStyler(open, close, this[STYLER]), this[IS_EMPTY]);
+        };
+        Object.defineProperty(this, styleName, { value: styleFunction });
+        return styleFunction;
+      }
+    };
+  }
+}
+var proto = Object.defineProperties(() => {}, {
+  ...styles2,
+  level: {
+    enumerable: true,
+    get() {
+      return this[GENERATOR].level;
+    },
+    set(level) {
+      this[GENERATOR].level = level;
+    }
+  }
+});
+var createStyler = (open, close, parent) => {
+  let openAll;
+  let closeAll;
+  if (parent === undefined) {
+    openAll = open;
+    closeAll = close;
+  } else {
+    openAll = parent.openAll + open;
+    closeAll = close + parent.closeAll;
+  }
+  return {
+    open,
+    close,
+    openAll,
+    closeAll,
+    parent
+  };
+};
+var createBuilder = (self, _styler, _isEmpty) => {
+  const builder = (...arguments_) => {
+    if (arguments_.length === 1) {
+      return applyStyle(builder, "" + arguments_[0]);
+    }
+    if (arguments_.length === 2) {
+      return applyStyle(builder, arguments_[0] + " " + arguments_[1]);
+    }
+    return applyStyle(builder, arguments_.join(" "));
+  };
+  Object.setPrototypeOf(builder, proto);
+  builder[GENERATOR] = self[GENERATOR] ?? self;
+  builder[STYLER] = _styler;
+  builder[IS_EMPTY] = _isEmpty;
+  return builder;
+};
+var applyStyle = (self, string) => {
+  if (self[GENERATOR][LEVEL] <= 0 || !string) {
+    return self[IS_EMPTY] ? "" : string;
+  }
+  let styler = self[STYLER];
+  if (styler === undefined) {
+    return string;
+  }
+  const { openAll, closeAll } = styler;
+  if (string.includes("\x1B")) {
+    while (styler !== undefined) {
+      string = stringReplaceAll(string, styler.close, styler.open);
+      styler = styler.parent;
+    }
+  }
+  const lfIndex = string.indexOf(`
+`);
+  if (lfIndex !== -1) {
+    string = stringEncaseCRLFWithFirstIndex(string, closeAll, openAll, lfIndex);
+  }
+  return openAll + string + closeAll;
+};
+Object.defineProperties(createChalk.prototype, { ...styles2, level: levelDescriptor });
+var chalk = createChalk();
+var chalkStderr = createChalk({ level: stderrColor ? stderrColor.level : 0 });
+var source_default = chalk;
+
+// src/diff-view.ts
 var CONTEXT_LINES = 3;
 var MAX_OUTPUT_LINES = 80;
 var SYNTAX_THEME = {
-  default: import_chalk.default.rgb(248, 248, 242),
-  keyword: import_chalk.default.rgb(249, 38, 114),
-  built_in: import_chalk.default.rgb(102, 217, 239),
-  type: import_chalk.default.rgb(102, 217, 239),
-  literal: import_chalk.default.rgb(174, 129, 255),
-  number: import_chalk.default.rgb(174, 129, 255),
-  regexp: import_chalk.default.rgb(230, 219, 116),
-  string: import_chalk.default.rgb(230, 219, 116),
-  subst: import_chalk.default.rgb(248, 248, 242),
-  symbol: import_chalk.default.rgb(174, 129, 255),
-  class: import_chalk.default.rgb(166, 226, 46),
-  function: import_chalk.default.rgb(166, 226, 46),
-  title: import_chalk.default.rgb(166, 226, 46),
-  params: import_chalk.default.rgb(253, 151, 31),
-  comment: import_chalk.default.rgb(117, 113, 94),
-  doctag: import_chalk.default.rgb(117, 113, 94),
-  meta: import_chalk.default.rgb(249, 38, 114),
-  "meta-keyword": import_chalk.default.rgb(249, 38, 114),
-  "meta-string": import_chalk.default.rgb(230, 219, 116),
-  section: import_chalk.default.rgb(166, 226, 46),
-  tag: import_chalk.default.rgb(249, 38, 114),
-  name: import_chalk.default.rgb(249, 38, 114),
-  attr: import_chalk.default.rgb(166, 226, 46),
-  attribute: import_chalk.default.rgb(166, 226, 46),
-  variable: import_chalk.default.rgb(248, 248, 242),
-  bullet: import_chalk.default.rgb(174, 129, 255),
-  code: import_chalk.default.rgb(230, 219, 116),
-  emphasis: import_chalk.default.italic,
-  strong: import_chalk.default.bold,
-  formula: import_chalk.default.rgb(248, 248, 242),
-  link: import_chalk.default.rgb(102, 217, 239).underline,
-  quote: import_chalk.default.rgb(117, 113, 94).italic,
-  addition: import_chalk.default.rgb(166, 226, 46),
-  deletion: import_chalk.default.rgb(249, 38, 114)
+  default: source_default.rgb(248, 248, 242),
+  keyword: source_default.rgb(249, 38, 114),
+  built_in: source_default.rgb(102, 217, 239),
+  type: source_default.rgb(102, 217, 239),
+  literal: source_default.rgb(174, 129, 255),
+  number: source_default.rgb(174, 129, 255),
+  regexp: source_default.rgb(230, 219, 116),
+  string: source_default.rgb(230, 219, 116),
+  subst: source_default.rgb(248, 248, 242),
+  symbol: source_default.rgb(174, 129, 255),
+  class: source_default.rgb(166, 226, 46),
+  function: source_default.rgb(166, 226, 46),
+  title: source_default.rgb(166, 226, 46),
+  params: source_default.rgb(253, 151, 31),
+  comment: source_default.rgb(117, 113, 94),
+  doctag: source_default.rgb(117, 113, 94),
+  meta: source_default.rgb(249, 38, 114),
+  "meta-keyword": source_default.rgb(249, 38, 114),
+  "meta-string": source_default.rgb(230, 219, 116),
+  section: source_default.rgb(166, 226, 46),
+  tag: source_default.rgb(249, 38, 114),
+  name: source_default.rgb(249, 38, 114),
+  attr: source_default.rgb(166, 226, 46),
+  attribute: source_default.rgb(166, 226, 46),
+  variable: source_default.rgb(248, 248, 242),
+  bullet: source_default.rgb(174, 129, 255),
+  code: source_default.rgb(230, 219, 116),
+  emphasis: source_default.italic,
+  strong: source_default.bold,
+  formula: source_default.rgb(248, 248, 242),
+  link: source_default.rgb(102, 217, 239).underline,
+  quote: source_default.rgb(117, 113, 94).italic,
+  addition: source_default.rgb(166, 226, 46),
+  deletion: source_default.rgb(249, 38, 114)
 };
 function patchResets(text, bg) {
   return text.replace(/\x1b\[0m/g, `\x1B[0m${bg}`);
@@ -46268,13 +46875,13 @@ async function listAgentsCommand(api, opts) {
 // src/agent-command.ts
 function printAgentHelp() {
   process.stderr.write(`
-todoforai-cli agent \u2014 inspect and update agent settings
+tfa-cli agent \u2014 inspect and update agent settings
 
 Usage:
-  todoforai-cli agent list                            List agents (name, model, id, paths)
-  todoforai-cli agent get <agent>                     Show a single agent's settings
-  todoforai-cli agent update <agent> <field=value>\u2026   Update one or more settings
-  todoforai-cli agent delete <agent>                  Delete an agent configuration (asks to confirm)
+  tfa-cli agent list                            List agents (name, model, id, paths)
+  tfa-cli agent get <agent>                     Show a single agent's settings
+  tfa-cli agent update <agent> <field=value>\u2026   Update one or more settings
+  tfa-cli agent delete <agent>                  Delete an agent configuration (asks to confirm)
 
 <agent> is a name or id (unique partial name also works).
 Fields map directly to agent settings; values are parsed as JSON when possible
@@ -46287,9 +46894,9 @@ Fields map directly to agent settings; values are parsed as JSON when possible
   name            agent display name
 
 Examples:
-  todoforai-cli agent update <agent> model=claude
-  todoforai-cli agent update <agent> model=anthropic:anthropic/claude-opus-5 temperature=0.5
-  todoforai-cli agent update <agent> sysmsg="You are a terse video editor."
+  tfa-cli agent update <agent> model=claude
+  tfa-cli agent update <agent> model=anthropic:anthropic/claude-opus-5 temperature=0.5
+  tfa-cli agent update <agent> sysmsg="You are a terse video editor."
 `);
 }
 var FIELD_ALIASES = {
@@ -46357,7 +46964,7 @@ async function agentCommand(api, positionals, args, formatPath) {
     const query = positionals[2];
     const assignments = positionals.slice(3);
     if (!query || !assignments.length) {
-      process.stderr.write(`${RED}Usage: todoforai-cli agent update <name|id> <field=value>\u2026${RESET}
+      process.stderr.write(`${RED}Usage: tfa-cli agent update <name|id> <field=value>\u2026${RESET}
 `);
       process.exit(2);
     }
@@ -46378,7 +46985,7 @@ async function agentCommand(api, positionals, args, formatPath) {
   if (sub === "delete" || sub === "rm") {
     const query = positionals[2];
     if (!query) {
-      process.stderr.write(`${RED}Usage: todoforai-cli agent delete <name|id>${RESET}
+      process.stderr.write(`${RED}Usage: tfa-cli agent delete <name|id>${RESET}
 `);
       process.exit(2);
     }
@@ -46601,6 +47208,87 @@ async function listTodosCommand(api, defaultProjectId, argv) {
 `);
 }
 
+// src/isolated.ts
+import { spawn as spawn2 } from "child_process";
+async function mintMayflyToken(apiUrl, apiKey, todoId) {
+  const res = await fetch(`${apiUrl}${restBasePath(apiKey)}/cli/mayfly/token`, {
+    method: "POST",
+    headers: { "content-type": "application/json", "x-api-key": apiKey },
+    body: JSON.stringify({ todoId }),
+    signal: AbortSignal.timeout(30000)
+  });
+  const text = await res.text();
+  if (!res.ok)
+    throw new Error(`--isolated: could not mint a session token (${res.status} ${text.slice(0, 200)}).`);
+  const token = JSON.parse(text)?.token;
+  if (typeof token !== "string" || !token)
+    throw new Error("--isolated: token mint returned no token.");
+  return token;
+}
+async function spawnMayflyBridge(apiUrl, todoId, workspace, opts = {}) {
+  if (!hasBridge())
+    throw new Error("--isolated requires `todoforai-bridge` on PATH.");
+  const loginless = !!opts.apiKey && !opts.apiKey.startsWith("dst_");
+  if (!loginless && !ensureBridgeCredentials(apiUrl))
+    throw new Error("`todoforai-bridge login` did not complete.");
+  const sessionToken = loginless ? await mintMayflyToken(apiUrl, opts.apiKey, todoId) : null;
+  const args = [...bridgeRunArgs(apiUrl), "--mayfly", todoId, "--workspace", workspace];
+  const env2 = { ...process.env };
+  if (sessionToken) {
+    for (const n of ["API_TOKEN", "MAYFLY_API_KEY"])
+      for (const p of ["TODOFORAI_", "TODO4AI_"])
+        delete env2[p + n];
+    env2.TODOFORAI_MAYFLY_TOKEN = sessionToken;
+  }
+  const child = spawn2("todoforai-bridge", args, { stdio: ["ignore", "pipe", "pipe"], env: env2 });
+  let stopped = false;
+  const stop = () => {
+    if (stopped)
+      return;
+    stopped = true;
+    try {
+      child.kill("SIGTERM");
+    } catch {}
+  };
+  process.on("exit", stop);
+  const readyRx = new RegExp(`^BRIDGE_READY id=mayfly-${todoId}\\b`, "m");
+  await new Promise((resolvePromise, reject) => {
+    const timeoutMs = opts.timeoutMs ?? 15000;
+    let buf = "";
+    let lastErr = "";
+    const timer = setTimeout(() => {
+      stop();
+      reject(new Error(`Isolated bridge not ready after ${timeoutMs / 1000}s.${lastErr ? ` ${lastErr}` : ""}`));
+    }, timeoutMs);
+    child.stdout.on("data", (d) => {
+      buf += d.toString();
+      if (readyRx.test(buf)) {
+        clearTimeout(timer);
+        resolvePromise();
+      }
+    });
+    child.stderr.on("data", (d) => {
+      const s = d.toString();
+      const err = s.match(/^error: .*$/m);
+      if (err)
+        lastErr = err[0];
+      if (opts.debug)
+        process.stderr.write(d);
+    });
+    child.on("error", (err) => {
+      clearTimeout(timer);
+      reject(err);
+    });
+    child.on("exit", (code) => {
+      if (!stopped) {
+        clearTimeout(timer);
+        reject(new Error(lastErr || `Isolated bridge exited early (code ${code}).`));
+      }
+    });
+  });
+  return { child, stop };
+}
+
 // src/index.ts
 try {
   const pkgPath = path3.resolve(fileURLToPath(import.meta.url), "../../package.json");
@@ -46689,7 +47377,7 @@ Cancelled by user (Ctrl+C)
   const groupName = args["group-name"] === undefined ? undefined : String(args["group-name"]);
   if (positionals[0] === "start") {
     if (!positionals[1] && !args.template) {
-      process.stderr.write(`${RED}Usage: todoforai-cli start <todo-id>${RESET}
+      process.stderr.write(`${RED}Usage: tfa-cli start <todo-id>${RESET}
 `);
       process.exit(2);
     }
@@ -46716,6 +47404,23 @@ Cancelled by user (Ctrl+C)
     printUsage();
     process.exit(0);
   }
+  if (args.isolated) {
+    if (args.resume || args.continue) {
+      process.stderr.write(`Error: --isolated cannot be combined with --resume/--continue (isolation binds at create)
+`);
+      process.exit(2);
+    }
+    if (args["no-watch"]) {
+      process.stderr.write(`Error: --isolated requires watching the run (drop --no-watch)
+`);
+      process.exit(2);
+    }
+    if (args.template) {
+      process.stderr.write(`Error: --isolated does not support --template yet
+`);
+      process.exit(2);
+    }
+  }
   const cfg = new ConfigStore(args["config-path"]);
   if (args["show-config"]) {
     console.log(`Config file: ${formatPathWithTilde(cfg.path)}`);
@@ -46756,12 +47461,12 @@ Cancelled by user (Ctrl+C)
 
 `);
     try {
-      const { spawn: spawn2 } = await import("child_process");
+      const { spawn: spawn3 } = await import("child_process");
       if (process.platform === "win32") {
-        spawn2("cmd", ["/c", "start", "", url], { stdio: "ignore", detached: true }).unref();
+        spawn3("cmd", ["/c", "start", "", url], { stdio: "ignore", detached: true }).unref();
       } else {
         const cmd = process.platform === "darwin" ? "open" : "xdg-open";
-        spawn2(cmd, [url], { stdio: "ignore", detached: true }).unref();
+        spawn3(cmd, [url], { stdio: "ignore", detached: true }).unref();
       }
     } catch {}
     process.stderr.write(`Waiting for approval (expires in ${Math.round(expiresIn / 60)}min)...
@@ -46801,6 +47506,20 @@ Cancelled by user (Ctrl+C)
   if (!apiKey) {
     apiKey = await deviceLogin();
   }
+  if (args["debug-dump"]) {
+    const debugSecret = getEnv("DEBUG_SECRET");
+    if (!debugSecret) {
+      process.stderr.write(`Error: --debug-dump requires TODOFORAI_DEBUG_SECRET
+`);
+      process.exit(2);
+    }
+    let extraHeaders = {};
+    try {
+      extraHeaders = JSON.parse(process.env.TODOFORAI_EXTRA_HEADERS || "{}");
+    } catch {}
+    extraHeaders["x-tfa-debug"] = debugSecret;
+    process.env.TODOFORAI_EXTRA_HEADERS = JSON.stringify(extraHeaders);
+  }
   const api = new ApiClient(apiUrl, apiKey, args["user-id"]);
   if (args["user-id"] && args.safe) {
     process.stderr.write(`Error: --safe is not supported with --user-id
@@ -46831,7 +47550,7 @@ Cancelled by user (Ctrl+C)
   if (positionals[0] === "delete") {
     const todoId = positionals[1];
     if (!todoId) {
-      process.stderr.write(`${RED}Usage: todoforai-cli delete <todo-id>${RESET}
+      process.stderr.write(`${RED}Usage: tfa-cli delete <todo-id>${RESET}
 `);
       process.exit(2);
     }
@@ -46844,7 +47563,7 @@ Cancelled by user (Ctrl+C)
     const [, todoId, ...rest] = positionals;
     const content2 = rest.join(" ") || await readStdin();
     if (!todoId || !content2) {
-      process.stderr.write(`${RED}Usage: todoforai-cli addmessage <todo-id> "content"${RESET}
+      process.stderr.write(`${RED}Usage: tfa-cli addmessage <todo-id> "content"${RESET}
 `);
       process.exit(2);
     }
@@ -46863,7 +47582,7 @@ Cancelled by user (Ctrl+C)
     const todoId = positionals[2] || (!explicitProject ? getEnv("TODO_ID") || cfgScope.data.last_todo_id : undefined);
     const projectId2 = todoId ? undefined : args.project || getEnv("PROJECT_ID") || cfgScope.data.default_project_id;
     if (!todoId && !projectId2) {
-      process.stderr.write(`${RED}Usage: todoforai-cli show list [todo-id] [--project <id>] [--card <name>]${RESET}
+      process.stderr.write(`${RED}Usage: tfa-cli show list [todo-id] [--project <id>] [--card <name>]${RESET}
 `);
       process.exit(2);
     }
@@ -46888,7 +47607,7 @@ Cancelled by user (Ctrl+C)
     const [, filePath, todoArg] = positionals;
     const todoId = todoArg || getEnv("TODO_ID") || cfgScope.data.last_todo_id;
     if (!filePath || !todoId) {
-      process.stderr.write(`${RED}Usage: todoforai-cli show <file|-> [todo-id]${RESET}
+      process.stderr.write(`${RED}Usage: tfa-cli show <file|-> [todo-id]${RESET}
 `);
       process.exit(2);
     }
@@ -46922,7 +47641,7 @@ Cancelled by user (Ctrl+C)
     const [, url, todoArg] = positionals;
     const todoId = todoArg || getEnv("TODO_ID") || cfgScope.data.last_todo_id;
     if (!url || !todoId) {
-      process.stderr.write(`${RED}Usage: todoforai-cli open <url> [todo-id]${RESET}
+      process.stderr.write(`${RED}Usage: tfa-cli open <url> [todo-id]${RESET}
 `);
       process.exit(2);
     }
@@ -46936,7 +47655,7 @@ Cancelled by user (Ctrl+C)
   if (positionals[0] === "recommend") {
     const templateId = args.template || positionals[1];
     if (!templateId) {
-      process.stderr.write(`${RED}Usage: todoforai-cli recommend --template <id> [--note "why"] [--priority high|medium|low] [--title "..."] [--group <slug> [--group-name "..."] [--group-description "..."]] [--project <id>]${RESET}
+      process.stderr.write(`${RED}Usage: tfa-cli recommend --template <id> [--note "why"] [--priority high|medium|low] [--title "..."] [--group <slug> [--group-name "..."] [--group-description "..."]] [--project <id>]${RESET}
 `);
       process.stderr.write(`${DIM}Create a template first with: todoregistry-cli create --name ... --description ... --body @prompt.md${RESET}
 `);
@@ -46983,7 +47702,7 @@ Cancelled by user (Ctrl+C)
       seedProjectId = projects2.find((p) => p.project?.isDefault)?.project?.id || projects2[0]?.project?.id;
     }
     if (!seedProjectId) {
-      process.stderr.write(`${RED}Usage: todoforai-cli claim mint --seed <projectId> [--emails a@x,b@y] [--ttl <sec>]${RESET}
+      process.stderr.write(`${RED}Usage: tfa-cli claim mint --seed <projectId> [--emails a@x,b@y] [--ttl <sec>]${RESET}
 `);
       process.exit(2);
     }
@@ -47226,7 +47945,12 @@ Resumed: ${CYAN}${getFrontendUrl(apiUrl, todoId)}${RESET}
   }
   const ws = args["no-watch"] ? null : new FrontendWebSocket(apiUrl, apiKey);
   const wsReady = ws ? ws.connect() : null;
-  const bridgeReady = !args["no-bridge"] && !args["no-watch"] ? ensureBridgeRunning(apiUrl, apiKey) : null;
+  const bridgeReady = !args["no-bridge"] && !args["no-watch"] && !args.isolated ? ensureBridgeRunning(apiUrl, apiKey) : null;
+  const isolatedTodoId = args.isolated ? crypto.randomUUID() : undefined;
+  const mayfly = isolatedTodoId ? await spawnMayflyBridge(apiUrl, isolatedTodoId, realpathSync(resolve3(args.path || ".")), { debug: !!args.debug, apiKey }) : null;
+  if (mayfly)
+    process.stderr.write(`${DIM}Isolated bridge:${RESET} ${CYAN}mayfly-${isolatedTodoId}${RESET}
+`);
   let preMatchedAgent = null;
   let agents = null;
   if (args.agent) {
@@ -47337,7 +48061,7 @@ Resumed: ${CYAN}${getFrontendUrl(apiUrl, todoId)}${RESET}
   cfg.addToHistory(content);
   let todo;
   try {
-    todo = await api.addMessage(projectId, content, agent, undefined, undefined, undefined, groupTag || undefined, groupName);
+    todo = await api.addMessage(projectId, content, agent, isolatedTodoId, undefined, undefined, groupTag || undefined, groupName);
   } catch (e) {
     if (!args.project && !envProjectId && cfgScope.data.default_project_id === projectId && /failed: 403/.test(e.message || "")) {
       cfgScope.clearDefaultProject();
@@ -47373,9 +48097,18 @@ ${"\u2500".repeat(40)}
     }
     await ws.close();
   }
+  mayfly?.stop();
 }
-main().then(() => process.exit(process.exitCode ?? 0)).catch((e) => {
+var finish = (code) => {
+  process.exitCode = code;
+  const poll = setInterval(() => {
+    if (process.stdout.writableLength === 0)
+      process.exit(code);
+  }, 5);
+  poll.unref?.();
+};
+main().then(() => finish(process.exitCode ?? 0)).catch((e) => {
   process.stderr.write(`Error: ${e.message}
 `);
-  process.exit(1);
+  finish(1);
 });