@doist/todoist-mcp 12.5.5 → 12.5.6

package/dist/index.js

@@ -1,6 +1,6 @@
-import { t as s } from "./mcp-server-Cm72B6O_.js";
-import { F as r, a as d, b as i, c as n, d as c, e as l, f as m, h as p, i as f, j as T, k as j, l as g, m as k, n as u, o as v, p as P, q as h, r as b, s as S, u as y, v as A, w as C, x as R, y as x, z as F, g as M, A as O, B as E, C as L, D as w, E as H, G as I, H as W, I as q, J as z, K as B, L as D, M as N, N as U, O as V, P as G, Q as J, R as K, S as Q, T as X, U as Y, V as _, W as Z, X as $, Y as ss } from "./mcp-server-Cm72B6O_.js";
-import { r as es, v as ts } from "./require-valid-todoist-token-DFE6h87D.js";
+import { t as s } from "./mcp-server-Drv9wBc-.js";
+import { F as r, a as d, b as i, c as n, d as c, e as l, f as m, h as p, i as f, j as T, k as j, l as g, m as k, n as u, o as v, p as P, q as h, r as b, s as S, u as y, v as A, w as C, x as R, y as x, z as F, g as M, A as O, B as E, C as L, D as w, E as H, G as I, H as W, I as q, J as z, K as B, L as D, M as N, N as U, O as V, P as G, Q as J, R as K, S as Q, T as X, U as Y, V as _, W as Z, X as $, Y as ss } from "./mcp-server-Drv9wBc-.js";
+import { r as es, v as ts } from "./require-valid-todoist-token-CC6hWyEx.js";
 const e = s;
 export {
   r as FEATURE_NAMES,

package/dist/main-http.js

@@ -3,8 +3,8 @@ import { isIP as p } from "node:net";
 import O from "dotenv";
 import { StreamableHTTPServerTransport as S } from "@modelcontextprotocol/sdk/server/streamableHttp.js";
 import m from "express";
-import { g as v } from "./mcp-server-Cm72B6O_.js";
-import { r as w } from "./require-valid-todoist-token-DFE6h87D.js";
+import { g as v } from "./mcp-server-Drv9wBc-.js";
+import { r as w } from "./require-valid-todoist-token-CC6hWyEx.js";
 function f(t) {
   return t.startsWith("[") && t.endsWith("]") ? t.slice(1, -1) : t;
 }

package/dist/main.js

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import { StdioServerTransport as s } from "@modelcontextprotocol/sdk/server/stdio.js";
 import c from "dotenv";
-import { g as i } from "./mcp-server-Cm72B6O_.js";
+import { g as i } from "./mcp-server-Drv9wBc-.js";
 function p() {
   const e = process.env.TODOIST_BASE_URL, r = process.env.TODOIST_API_KEY;
   if (!r)

package/dist/{mcp-server-Cm72B6O_.js → mcp-server-Drv9wBc-.js}

@@ -28,7 +28,7 @@ function zs() {
     }
   return console.error("Failed to load task list app HTML from any known path:", e), Bs;
 }
-const Ht = zs(), Ks = Rt("sha256").update(Ht).digest("hex").slice(0, 12), Le = `ui://todoist/task-list@${Ks}`, qs = "ui://todoist/task-list@{hash}", Ne = "Interactive task list widget", Ys = "https://ai.todoist.net", Me = {
+const Ht = zs(), qs = Rt("sha256").update(Ht).digest("hex").slice(0, 12), Le = `ui://todoist/task-list@${qs}`, Ks = "ui://todoist/task-list@{hash}", Ne = "Interactive task list widget", Ys = "https://ai.todoist.net", Me = {
   ui: {
     prefersBorder: !0,
     csp: {
@@ -69,7 +69,7 @@ function Gs(e) {
     async () => pt(Le)
   ), e.registerResource(
     "todoist-task-list-hash-fallback",
-    new Ds(qs, { list: void 0 }),
+    new Ds(Ks, { list: void 0 }),
     {
       description: `${Ne} compatibility fallback`,
       mimeType: Lt,
@@ -371,7 +371,7 @@ function or(e) {
   const t = Ve(e);
   return t ? sr(t) : Bt(e);
 }
-const nr = "12.5.5", ir = {
+const nr = "12.5.6", ir = {
   version: nr
 }, C = {
   /** Default limit for task listings */
@@ -407,7 +407,7 @@ const nr = "12.5.5", ir = {
 }, fe = {
   /** Maximum number of failures to show in detailed error messages */
   MAX_FAILURES_SHOWN: 3
-}, Kt = {
+}, qt = {
   /** Maximum tasks accepted by one task create or update operation */
   TASKS_PER_OPERATION: 25
 }, Be = {
@@ -475,7 +475,7 @@ function mt(e, { queueTimeoutMs: t = Be.QUEUE_WAIT_MS } = {}) {
     }
   };
 }
-function qt() {
+function Kt() {
   return {
     moves: mt(Be.TASK_MOVES),
     writes: mt(Be.WRITES)
@@ -484,7 +484,7 @@ function qt() {
 const ht = /* @__PURE__ */ new Map(), Yt = /* @__PURE__ */ new WeakMap();
 let ft;
 function ar() {
-  return ft ??= qt(), ft;
+  return ft ??= Kt(), ft;
 }
 function cr(e) {
   return Rt("sha256").update(e).digest("hex");
@@ -492,7 +492,7 @@ function cr(e) {
 function dr(e, t) {
   const r = cr(t);
   let o = ht.get(r);
-  o || (o = qt(), ht.set(r, o)), Yt.set(e, o);
+  o || (o = Kt(), ht.set(r, o)), Yt.set(e, o);
 }
 function Gt(e) {
   return Yt.get(e) ?? ar();
@@ -811,15 +811,15 @@ const Br = [
   g.FIND_PROJECT_COLLABORATORS,
   g.FIND_COMPLETED_TASKS
 ];
-function Ke(e) {
+function qe(e) {
   if (e == null)
     return e;
   if (Array.isArray(e))
-    return e.map((t) => Ke(t));
+    return e.map((t) => qe(t));
   if (typeof e == "object") {
     const t = {};
     for (const [r, o] of Object.entries(e))
-      r !== "email" && (t[r] = Ke(o));
+      r !== "email" && (t[r] = qe(o));
     return t;
   }
   return e;
@@ -828,7 +828,7 @@ function zr(e) {
   const t = /\s*\([^)]*@[^)]+\)/g, r = /\S+@\S+\.\S+/g;
   return e.replace(t, "").replace(r, "[email hidden]");
 }
-function Kr({
+function qr({
   tool: e,
   server: t,
   client: r,
@@ -840,7 +840,7 @@ function Kr({
         e.name,
         () => O(() => i(d, r))
       );
-      return n && (u && (u = zr(u)), p && (p = Ke(p))), Ur({ textContent: u, structuredContent: p, contentItems: m });
+      return n && (u && (u = zr(u)), p && (p = qe(p))), Ur({ textContent: u, structuredContent: p, contentItems: m });
     } catch (u) {
       return console.error(`Error executing tool ${e.name}:`, { args: Mr(d), error: u }), Lr(or(u));
     }
@@ -865,7 +865,7 @@ function Kr({
   }
   t.registerTool(e.name, c, a);
 }
-const qr = {
+const Kr = {
   period: s.enum(["today", "7d", "14d", "30d", "this-week", "this-month"]).default("7d").describe(
     'Time period for the analysis. "today" analyzes the current day, "7d"/"14d"/"30d" analyze the last N days, "this-week" and "this-month" analyze the current week or month.'
   ),
@@ -1001,7 +1001,7 @@ const he = {
   name: "productivity-analysis",
   title: "Productivity Analysis",
   description: "Analyze your Todoist productivity with insights on completion trends, goal streaks, project distribution, and actionable recommendations. Gathers data from multiple tools and synthesizes a comprehensive report.",
-  argsSchema: qr,
+  argsSchema: Kr,
   callback: Jr
 };
 class _ extends Error {
@@ -1539,7 +1539,7 @@ const ue = s.enum(Ie).optional().catch(void 0), F = s.object({
   sectionOrder: s.number().describe("Ordering index within the project."),
   description: s.string().optional().describe("Supports Markdown.")
 });
-function K({ id: e, name: t, sectionOrder: r, description: o }) {
+function q({ id: e, name: t, sectionOrder: r, description: o }) {
   return { id: e, name: t, sectionOrder: r, description: o ?? void 0 };
 }
 const bo = s.object({
@@ -1883,7 +1883,7 @@ function Se(e, t = 5) {
   }
   return o;
 }
-function q({
+function K({
   subject: e,
   count: t,
   limit: r,
@@ -1978,9 +1978,9 @@ const as = new Bo(), zo = s.object({
   workspace: s.string().trim().min(1).optional().describe(
     "The workspace to create the project in. Accepts a workspace name or workspace ID. If not provided, creates a personal project. Use list-workspaces to see available workspaces."
   )
-}), Ko = {
+}), qo = {
   projects: s.array(zo).min(1).describe("The array of projects to add.")
-}, qo = {
+}, Ko = {
   projects: s.array(Q).describe("The created projects."),
   totalCount: s.number().describe("The total number of projects created."),
   failures: s.array(ee).describe(
@@ -1992,8 +1992,8 @@ const as = new Bo(), zo = s.object({
 }, Yo = {
   name: g.ADD_PROJECTS,
   description: "Add one or more new projects.",
-  parameters: Ko,
-  outputSchema: qo,
+  parameters: qo,
+  outputSchema: Ko,
   annotations: { readOnlyHint: !1, destructiveHint: !1, idempotentHint: !1 },
   async execute({ projects: e }, t) {
     const r = /* @__PURE__ */ new Map(), o = (l) => {
@@ -2185,7 +2185,7 @@ const rn = s.object({
     }), {
       textContent: cn({ sections: n, failures: i }),
       structuredContent: {
-        sections: n.map(K),
+        sections: n.map(q),
         totalCount: n.length,
         failures: i,
         totalRequested: e.length,
@@ -2426,7 +2426,7 @@ const at = new dn();
 function U(e) {
   return s.string().transform((t) => t === "" ? void 0 : t).optional().describe(e);
 }
-const Tt = Kt.TASKS_PER_OPERATION, ln = s.object({
+const Tt = qt.TASKS_PER_OPERATION, ln = s.object({
   content: s.string().min(1).describe(
     'The task name/title. Should be concise and actionable (e.g., "Review PR #123", "Call dentist"). For longer content, use the description field instead. Supports Markdown.'
   ),
@@ -2999,7 +2999,7 @@ const ps = ["task", "project", "comment", "section"], Rn = {
           const i = await t.getSection(o);
           if (!i)
             throw new Error(`Section ${o} not found.`);
-          const a = K(i);
+          const a = q(i);
           return {
             textContent: `Found section: ${a.name} • id=${a.id}`,
             structuredContent: {
@@ -3110,7 +3110,9 @@ const Wn = {
   ]).optional().describe("Type of event to filter by."),
   projectId: s.string().optional().describe("Filter events by parent project ID."),
   taskId: s.string().optional().describe("Filter events by parent task ID (for subtask events)."),
-  initiatorId: s.string().optional().describe("Filter by the user ID who initiated the event."),
+  initiatorId: s.string().optional().describe(
+    'Filter by the user ID who initiated the event. A first-person question ("what did I complete") needs this: without it the results cover every collaborator. Get the ID from user-info when you do not already have it.'
+  ),
   dateFrom: St(
     `Inclusive start of the activity range, as an ISO 8601 date or date-time. For all events on one local calendar day, use that day's start, for example "2026-08-02T00:00:00-04:00". Natural-language dates such as "tomorrow" are not supported.`
   ),
@@ -3127,7 +3129,7 @@ const Wn = {
   appliedFilters: s.record(s.string(), s.unknown())
 }, zn = {
   name: g.FIND_ACTIVITY,
-  description: 'Retrieve activity logs to monitor and audit changes in Todoist. Shows events from all users by default (use initiatorId to filter by specific user). To answer what someone completed in a period, including recurring task occurrences, use objectType "task", eventType "completed", and dateFrom/dateTo. Track task completions, updates, deletions, project changes, and more with flexible filtering. Activity history availability and retention depend on the user plan.',
+  description: `Retrieve activity logs to monitor and audit changes in Todoist. Shows events from all users by default (use initiatorId to filter by specific user). To answer what someone completed in a period, including recurring task occurrences, use objectType "task", eventType "completed", and dateFrom/dateTo. For a first-person question ("what did I get done"), also set initiatorId to the current user from user-info, or the answer includes collaborators' completions. Track task completions, updates, deletions, project changes, and more with flexible filtering. Activity history availability and retention depend on the user plan.`,
   parameters: Wn,
   outputSchema: Bn,
   annotations: { readOnlyHint: !0, destructiveHint: !1, idempotentHint: !0 },
@@ -3150,7 +3152,7 @@ const Wn = {
     r && n ? m.objectEventTypes = `${r}:${n}` : r ? m.objectEventTypes = `${r}:` : n && (m.objectEventTypes = `:${n}`), o && o !== "remove" && (m.objectId = o), i && (m.parentProjectId = i), a && (m.parentItemId = a), c && (m.initiatorId = c), d && (m.dateFrom = d), l && (m.dateTo = l);
     const { results: h, nextCursor: y } = await t.getActivityLogs(m), f = h.map(po);
     return {
-      textContent: Kn({ events: f, args: e, nextCursor: y }),
+      textContent: qn({ events: f, args: e, nextCursor: y }),
       structuredContent: {
         events: f,
         nextCursor: y ?? void 0,
@@ -3161,7 +3163,7 @@ const Wn = {
     };
   }
 };
-function Kn({
+function qn({
   events: e,
   args: t,
   nextCursor: r
@@ -3176,17 +3178,17 @@ function Kn({
   const i = [];
   t.objectId && i.push(`object ID: ${t.objectId}`), t.projectId && i.push(`project: ${t.projectId}`), t.taskId && i.push(`task: ${t.taskId}`), t.initiatorId && i.push(`initiator: ${t.initiatorId}`), t.dateFrom && i.push(`from: ${t.dateFrom}`), t.dateTo && i.push(`until (exclusive): ${t.dateTo}`);
   const a = [];
-  return e.length === 0 && (a.push("No activity events match the specified filters"), a.push("Note: Activity logs only show recent events"), t.eventType && a.push(`Try removing the eventType filter (${t.eventType})`), t.objectType && a.push(`Try removing the objectType filter (${t.objectType})`), (t.objectId || t.projectId || t.taskId) && a.push("Verify the object ID is correct")), q({
+  return e.length === 0 && (a.push("No activity events match the specified filters"), a.push("Note: Activity logs only show recent events"), t.eventType && a.push(`Try removing the eventType filter (${t.eventType})`), t.objectType && a.push(`Try removing the objectType filter (${t.objectType})`), (t.objectId || t.projectId || t.taskId) && a.push("Verify the object ID is correct")), K({
     subject: o,
     count: e.length,
     limit: t.limit,
     nextCursor: r ?? void 0,
     filterHints: i,
-    previewLines: qn(e, Math.min(e.length, t.limit)),
+    previewLines: Kn(e, Math.min(e.length, t.limit)),
     zeroReasonHints: a
   });
 }
-function qn(e, t = 10) {
+function Kn(e, t = 10) {
   const o = e.slice(0, t).map(Yn).join(`
 `);
   if (e.length > t) {
@@ -3397,7 +3399,7 @@ const oi = {
   appliedFilters: s.record(s.string(), s.unknown())
 }, ii = {
   name: g.FIND_COMPLETED_TASKS,
-  description: "Get completed tasks. since/until are optional and default to a 7-day window when omitted. Includes all collaborators by default. Person-specific queries (summaries, plans, reports) require responsibleUser.",
+  description: 'Get completed tasks in a date range. For "what did I complete/get done" questions, use find-activity instead — it reports completion events, including every occurrence of a recurring task, which this tool does not. since/until are optional and default to a 7-day window when omitted. Includes all collaborators by default. Person-specific queries (summaries, plans, reports) require responsibleUser.',
   parameters: oi,
   outputSchema: ni,
   annotations: { readOnlyHint: !0, destructiveHint: !1, idempotentHint: !0 },
@@ -3467,7 +3469,7 @@ function ai({
     a.push(`assigned to: ${d}`);
   }
   const c = [];
-  return e.length === 0 && (c.push("No tasks completed in this date range"), c.push("Try expanding the date range"), (t.projectId || t.sectionId || t.parentId) && c.push("Try removing project/section/parent filters"), t.getBy === "due" && c.push('Try switching to "completion" date instead')), q({
+  return e.length === 0 && (c.push("No tasks completed in this date range"), c.push("Try expanding the date range"), (t.projectId || t.sectionId || t.parentId) && c.push("Try removing project/section/parent filters"), t.getBy === "due" && c.push('Try switching to "completion" date instead')), K({
     subject: i,
     count: e.length,
     limit: t.limit,
@@ -3536,7 +3538,7 @@ function ui({
   const m = o.length > 0 ? `
 Shared labels (${o.length}): ${o.join(", ")}` : `
 No shared labels.`;
-  return q({
+  return K({
     subject: n,
     count: e.length,
     limit: t.searchText ? void 0 : t.limit,
@@ -3716,7 +3718,7 @@ function ys({
   const l = [];
   e.length === 0 && r && (l.push(`No users match "${r}"`), l.push("Try a broader search term or check spelling"), o > 0 && l.push(`${o} users available without filter`));
   const u = [];
-  return e.length > 0 ? (u.push(`Use ${fs} with responsibleUser to assign new tasks`), u.push(`Use ${bs} with responsibleUser to reassign existing tasks`), u.push("Use user names, emails, or IDs for assignments")) : (u.push(`Use ${hs} to find other projects`), r && o > 0 && u.push("Try searching without filters to see all users")), q({
+  return e.length > 0 ? (u.push(`Use ${fs} with responsibleUser to assign new tasks`), u.push(`Use ${bs} with responsibleUser to reassign existing tasks`), u.push("Use user names, emails, or IDs for assignments")) : (u.push(`Use ${hs} to find other projects`), r && o > 0 && u.push("Try searching without filters to see all users")), K({
     subject: a,
     count: e.length,
     filterHints: c,
@@ -3792,7 +3794,7 @@ function wi({
   const a = 10, d = e.slice(0, a).map(Fo).join(`
 `), l = e.length - a, u = l > 0 ? `${d}
     …and ${l} more` : d, p = [];
-  return e.length === 0 && (t.searchText ? (p.push("Try broader search terms"), p.push("Check spelling"), p.push("Remove searchText to see all projects")) : t.archivedStatus === "archived" ? (p.push("No archived projects"), p.push("Use archivedStatus: 'all' to also see active projects")) : t.archivedStatus === "all" ? (p.push("No projects created yet"), p.push(`Use ${Ct} to create a project`)) : (p.push("No active projects"), p.push("Use archivedStatus: 'all' to include archived projects"), p.push(`Use ${Ct} to create a project`))), q({
+  return e.length === 0 && (t.searchText ? (p.push("Try broader search terms"), p.push("Check spelling"), p.push("Remove searchText to see all projects")) : t.archivedStatus === "archived" ? (p.push("No archived projects"), p.push("Use archivedStatus: 'all' to also see active projects")) : t.archivedStatus === "all" ? (p.push("No projects created yet"), p.push(`Use ${Ct} to create a project`)) : (p.push("No active projects"), p.push("Use archivedStatus: 'all' to include archived projects"), p.push(`Use ${Ct} to create a project`))), K({
     subject: n,
     count: e.length,
     limit: t.searchText || t.archivedStatus === "all" ? void 0 : t.limit,
@@ -3915,7 +3917,7 @@ const { ADD_SECTIONS: Ci } = g, $i = {
     e.searchText ? o = await uo(t, e.searchText, r) : o = (await t.getSections({
       projectId: r
     })).results;
-    const n = o.map(K);
+    const n = o.map(q);
     return {
       textContent: Ai({
         sections: n,
@@ -3939,7 +3941,7 @@ function Ai({
   r ? (o.push("Try broader search terms"), o.push("Check spelling"), o.push("Remove searchText to see all sections")) : (o.push("Project has no sections yet"), o.push(`Use ${Ci} to create sections`));
   const n = r ? `Sections in project ${t} matching "${r}"` : `Sections in project ${t}`, i = e.length > 0 ? e.map((a) => `    ${a.name} • id=${a.id}`).join(`
 `) : void 0;
-  return q({
+  return K({
     subject: n,
     count: e.length,
     previewLines: i,
@@ -4067,7 +4069,7 @@ function _i({
       a.push(`Great job! No tasks for today${c}`);
     } else
       a.push("Expand date range with larger 'daysCount'"), a.push("Check today's tasks with startDate='today'");
-  return q({
+  return K({
     subject: i,
     count: e.length,
     limit: t.limit,
@@ -4353,7 +4355,7 @@ function Ee({
       t.searchText && (c.push("Try broader search terms"), c.push("Verify spelling and try partial words"), t.responsibleUser || c.push(`Check completed tasks with ${$t}`));
     }
   }
-  return q({
+  return K({
     subject: i,
     count: e.length,
     limit: t.limit,
@@ -4441,7 +4443,7 @@ function zi(e) {
   }
   return o(r), r;
 }
-async function Ki(e, t) {
+async function qi(e, t) {
   const r = {};
   return await Promise.all(
     t.map(async (o) => {
@@ -4477,11 +4479,11 @@ function xt(e) {
   }
   return r;
 }
-function qe(e, t = "") {
+function Ke(e, t = "") {
   const r = [];
   for (const o of e) {
     const n = `id=${o.id}`, i = o.dueDate ? `; due=${o.dueDate}` : "", a = `; content=${o.content}`;
-    r.push(`${t}- ${n}${i}${a}`), o.children.length > 0 && r.push(...qe(o.children, `${t}  `));
+    r.push(`${t}- ${n}${i}${a}`), o.children.length > 0 && r.push(...Ke(o.children, `${t}  `));
   }
   return r;
 }
@@ -4492,13 +4494,13 @@ function vs(e, t) {
     parentId: B(e) ? e.parentId ?? void 0 : void 0,
     folderId: ke(e) ? e.folderId ?? void 0 : void 0,
     childOrder: e.childOrder,
-    sections: (t[e.id] || []).map(K),
+    sections: (t[e.id] || []).map(q),
     children: e.children.map(
       (r) => vs(r, t)
     )
   };
 }
-async function qi(e, t) {
+async function Ki(e, t) {
   let r = [], o;
   do {
     const { results: n, nextCursor: i } = await e.getTasks({
@@ -4515,7 +4517,7 @@ async function Yi(e, t) {
   return r;
 }
 async function Gi(e) {
-  const { results: t } = await e.getProjects({}), r = t.find((p) => B(p) && p.inboxProject === !0), o = t.filter((p) => !B(p) || p.inboxProject !== !0), n = zi(o), i = t.map((p) => p.id), a = await Ki(e, i), c = ["# Personal Projects", ""];
+  const { results: t } = await e.getProjects({}), r = t.find((p) => B(p) && p.inboxProject === !0), o = t.filter((p) => !B(p) || p.inboxProject !== !0), n = zi(o), i = t.map((p) => p.id), a = await qi(e, i), c = ["# Personal Projects", ""];
   if (r) {
     c.push(`- Inbox Project: ${r.name} (id=${r.id})`);
     for (const p of a[r.id] || [])
@@ -4538,7 +4540,7 @@ async function Gi(e) {
     inbox: r ? {
       id: r.id,
       name: r.name,
-      sections: (a[r.id] || []).map(K)
+      sections: (a[r.id] || []).map(q)
     } : void 0,
     projects: n.map(
       (p) => vs(p, a)
@@ -4556,7 +4558,7 @@ async function Vi(e, t) {
   const [r, o, n] = await Promise.all([
     e.getProject(t),
     Yi(e, t),
-    qi(e, t)
+    Ki(e, t)
   ]), i = {};
   for (const u of o)
     i[u.id] = [];
@@ -4567,7 +4569,7 @@ async function Vi(e, t) {
   if (a.length > 0) {
     c.push("");
     const u = xt(a);
-    c.push(...qe(u));
+    c.push(...Ke(u));
   }
   for (const u of o) {
     c.push(""), c.push(`## ${u.name}`);
@@ -4575,7 +4577,7 @@ async function Vi(e, t) {
     if (!p?.length)
       continue;
     const m = xt(p);
-    c.push(...qe(m));
+    c.push(...Ke(m));
   }
   const d = c.join(`
 `), l = {
@@ -4584,7 +4586,7 @@ async function Vi(e, t) {
       id: r.id,
       name: r.name
     },
-    sections: o.map(K),
+    sections: o.map(q),
     tasks: n.map((u) => ({
       ...u,
       children: []
@@ -4669,7 +4671,7 @@ const Ji = {
   projectColors: s.record(s.string(), s.string()).describe("Map of project ID to color key.")
 }, Qi = {
   name: g.GET_PRODUCTIVITY_STATS,
-  description: "Get comprehensive productivity statistics including daily/weekly completion breakdowns, goal streaks (current, last, max), karma score and trends, and historical karma data. Useful for productivity analysis and tracking goal progress.",
+  description: 'Get comprehensive productivity statistics including daily/weekly completion breakdowns, goal streaks (current, last, max), karma score and trends, and historical karma data. Useful for productivity analysis and tracking goal progress. Reports counts, streaks and karma only, never the tasks themselves — for "what did I complete", use find-activity.',
   parameters: Xi,
   outputSchema: Zi,
   annotations: {
@@ -5057,7 +5059,7 @@ const ya = {
       projectId: e,
       file: r ?? "",
       fileName: "template.csv"
-    }), a = i.sections.map(K), c = i.tasks.map(P), d = i.comments.map(ge), l = i.projects.length + a.length + c.length + d.length;
+    }), a = i.sections.map(q), c = i.tasks.map(P), d = i.comments.map(ge), l = i.projects.length + a.length + c.length + d.length;
     return {
       textContent: wa({
         projectId: e,
@@ -5567,7 +5569,7 @@ const Pa = {
   tasks: s.array(F).describe("The rescheduled tasks."),
   totalCount: s.number().describe("The total number of tasks rescheduled."),
   rescheduledTaskIds: s.array(s.string()).describe("The IDs of the rescheduled tasks.")
-}, Ka = {
+}, qa = {
   name: g.RESCHEDULE_TASKS,
   description: "Reschedule tasks to new dates while preserving recurring schedules. Unlike update-tasks (which replaces the entire due string and can wipe recurrence), this tool changes only the date, keeping recurrence patterns intact. Use this when moving recurring tasks to a different date without altering their repeat pattern.",
   parameters: Ba,
@@ -5584,7 +5586,7 @@ const Pa = {
         throw new Error(
           `Task "${l.content}" (${l.id}) has no due date. Rescheduling requires an existing due date.`
         );
-      const u = qa(d.date, l.due);
+      const u = Ka(d.date, l.due);
       return V("item_update", {
         id: d.id,
         due: {
@@ -5615,7 +5617,7 @@ const Pa = {
     };
   }
 };
-function qa(e, t) {
+function Ka(e, t) {
   if (/^\d{4}-\d{2}-\d{2}$/.test(e) && t.datetime) {
     const o = t.datetime.substring(10);
     return e + o;
@@ -6171,7 +6173,7 @@ const Cc = s.object({
         sections: r
       }),
       structuredContent: {
-        sections: r.map(K),
+        sections: r.map(q),
         totalCount: r.length,
         updatedSectionIds: r.map((n) => n.id)
       }
@@ -6263,7 +6265,7 @@ const Wc = s.object({
   isUncompletable: s.boolean().optional().describe(
     "Whether this task should be uncompletable (organizational header). Tasks with isUncompletable: true appear as organizational headers and cannot be completed."
   )
-}), Bc = (e) => !!(e.projectId || e.sectionId || e.parentId), js = (e) => Object.keys(e).length > 0, zc = ["remove", "no date"], Kc = ["remove", "no date", "no deadline"], qc = "no date", Ye = Kt.TASKS_PER_OPERATION, Ot = {
+}), Bc = (e) => !!(e.projectId || e.sectionId || e.parentId), js = (e) => Object.keys(e).length > 0, zc = ["remove", "no date"], qc = ["remove", "no date", "no deadline"], Kc = "no date", Ye = qt.TASKS_PER_OPERATION, Ot = {
   tasks: s.array(Wc).min(1).max(Ye).describe(`The tasks to update (max ${Ye}).`)
 }, Yc = {
   tasks: s.array(F).describe("The updated tasks."),
@@ -6548,10 +6550,10 @@ async function ed({
   const j = Pt(
     d,
     zc,
-    qc
+    Kc
   );
   j !== void 0 && (v = { ...v, dueString: j });
-  const A = Pt(h, Kc, null);
+  const A = Pt(h, qc, null);
   if (A !== void 0 && (v = { ...v, deadlineDate: A }), l)
     try {
       const { minutes: k } = Xt(l);
@@ -6849,7 +6851,7 @@ const vd = {
   completeTasks: wn,
   uncompleteTasks: Za,
   updateTasks: Gc,
-  rescheduleTasks: Ka,
+  rescheduleTasks: qa,
   findTasks: Fi,
   findTasksByDate: Sd,
   findCompletedTasks: ii,
@@ -6930,7 +6932,7 @@ What each tool does and how to fill its parameters is in the tool's own descript
 
 **Finding things**
 
-- For "what did I complete?", use **find-activity** with objectType="task" and eventType="completed". It reports completion events, including each occurrence of a recurring task. **find-completed-tasks** lists completed tasks, which is not the same question.
+- For "what did I complete?", call **user-info** for the asker's user ID, then **find-activity** with objectType="task", eventType="completed", initiatorId, and dateFrom/dateTo. Without initiatorId the answer covers every collaborator. find-activity reports completion events, including each occurrence of a recurring task; **find-completed-tasks** lists completed tasks, which is not the same question.
 - To resolve a person's name or email to a user ID, or to answer "who is X?", use **find-project-collaborators** with just a searchTerm. It covers every shared project you can access plus yourself, so an empty result means they collaborate on none of them — not that they do not exist.
 - Before assigning work to someone, call **find-project-collaborators** again with the target projectId. Resolving an ID only proves they collaborate on *some* project you can see; assignment fails unless they collaborate on that one.
 - To find out whether a task hides subtasks, use **fetch-object** with includeChildren rather than a speculative **find-tasks** call.
@@ -6976,7 +6978,7 @@ function Hd({
   Gs(o);
   const i = { server: o, client: n, features: r };
   for (const a of Cd)
-    Kr({ tool: a, ...i });
+    qr({ tool: a, ...i });
   return o.registerPrompt(
     he.name,
     {
@@ -7002,7 +7004,7 @@ export {
   Na as K,
   Cd as L,
   Ha as M,
-  Ka as N,
+  qa as N,
   Va as O,
   Za as P,
   rc as Q,

package/dist/mcp-server.d.ts

@@ -1,6 +1,6 @@
 import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 import { FEATURE_NAMES, Feature, FeatureName, Features } from './mcp-helpers.js';
-export declare const instructions = "\n## Todoist Task and Project Management Tools\n\nYou have access to comprehensive Todoist management tools for personal productivity and team collaboration. Use these tools to help users manage tasks, projects, sections, comments, and assignments effectively.\n\n### Core Capabilities:\n- Create, update, complete, and search tasks with rich metadata (priorities, due dates, durations, assignments)\n- Manage projects and sections with flexible organization\n- Handle comments and collaboration features\n- Bulk assignment operations for team workflows\n- Get overviews and insights about workload and progress\n\n### Choosing between tools\n\nWhat each tool does and how to fill its parameters is in the tool's own description and input schema. This section covers only what those cannot: which tool to reach for, and how tools relate to each other.\n\n**Dates**\n\n- To move a task to a different date use **reschedule-tasks**, never **update-tasks**. update-tasks replaces the whole due string, which destroys recurrence on recurring tasks.\n- Never send a task's existing projectId, sectionId or parentId back to **update-tasks** \u2014 those fields are treated as a move.\n- All dates respect the user's timezone.\n\n**Finding things**\n\n- For \"what did I complete?\", use **find-activity** with objectType=\"task\" and eventType=\"completed\". It reports completion events, including each occurrence of a recurring task. **find-completed-tasks** lists completed tasks, which is not the same question.\n- To resolve a person's name or email to a user ID, or to answer \"who is X?\", use **find-project-collaborators** with just a searchTerm. It covers every shared project you can access plus yourself, so an empty result means they collaborate on none of them \u2014 not that they do not exist.\n- Before assigning work to someone, call **find-project-collaborators** again with the target projectId. Resolving an ID only proves they collaborate on *some* project you can see; assignment fails unless they collaborate on that one.\n- To find out whether a task hides subtasks, use **fetch-object** with includeChildren rather than a speculative **find-tasks** call.\n- Filter tasks by label **name**. Label IDs are only for **delete-object** and **update-labels**. Shared labels can be renamed but not recoloured, reordered or favourited.\n\n**Deleting and archiving**\n\n- **delete-object** removes every object type; there is no per-type delete tool. Reminders use type \"reminder\", location reminders \"location_reminder\".\n- A workspace project must be archived with **project-management** before it can be deleted. A personal project can be deleted directly.\n\n**Templates**\n\n- Templates cannot be listed through this server, so only use an ID or URL the user gave you. To start a new project from one, call **add-projects** first and import into it. Imports write immediately and cannot be undone.\n\n**Project health**\n\n- Health data may be stale \u2014 check the isStale flag, and call **analyze-project-health** to refresh it before reading again.\n\n**Batches**\n\n- Prefer a batch tool over one call per item.\n- Where a batch tool returns per-item `failures` alongside successes, one failure does not undo the rest: never retry the whole batch, re-send only the items whose failure reason is fixable. **reschedule-tasks** is the exception \u2014 it has no per-item failures and throws if any task fails, so retry it as a whole.\n\nUse the **productivity-analysis** prompt for a combined productivity review \u2014 it pulls together user-info, get-productivity-stats and find-completed-tasks.\n\nAlways provide clear, actionable task titles and descriptions. Use the overview tools to give users context about their workload and project status.\n";
+export declare const instructions = "\n## Todoist Task and Project Management Tools\n\nYou have access to comprehensive Todoist management tools for personal productivity and team collaboration. Use these tools to help users manage tasks, projects, sections, comments, and assignments effectively.\n\n### Core Capabilities:\n- Create, update, complete, and search tasks with rich metadata (priorities, due dates, durations, assignments)\n- Manage projects and sections with flexible organization\n- Handle comments and collaboration features\n- Bulk assignment operations for team workflows\n- Get overviews and insights about workload and progress\n\n### Choosing between tools\n\nWhat each tool does and how to fill its parameters is in the tool's own description and input schema. This section covers only what those cannot: which tool to reach for, and how tools relate to each other.\n\n**Dates**\n\n- To move a task to a different date use **reschedule-tasks**, never **update-tasks**. update-tasks replaces the whole due string, which destroys recurrence on recurring tasks.\n- Never send a task's existing projectId, sectionId or parentId back to **update-tasks** \u2014 those fields are treated as a move.\n- All dates respect the user's timezone.\n\n**Finding things**\n\n- For \"what did I complete?\", call **user-info** for the asker's user ID, then **find-activity** with objectType=\"task\", eventType=\"completed\", initiatorId, and dateFrom/dateTo. Without initiatorId the answer covers every collaborator. find-activity reports completion events, including each occurrence of a recurring task; **find-completed-tasks** lists completed tasks, which is not the same question.\n- To resolve a person's name or email to a user ID, or to answer \"who is X?\", use **find-project-collaborators** with just a searchTerm. It covers every shared project you can access plus yourself, so an empty result means they collaborate on none of them \u2014 not that they do not exist.\n- Before assigning work to someone, call **find-project-collaborators** again with the target projectId. Resolving an ID only proves they collaborate on *some* project you can see; assignment fails unless they collaborate on that one.\n- To find out whether a task hides subtasks, use **fetch-object** with includeChildren rather than a speculative **find-tasks** call.\n- Filter tasks by label **name**. Label IDs are only for **delete-object** and **update-labels**. Shared labels can be renamed but not recoloured, reordered or favourited.\n\n**Deleting and archiving**\n\n- **delete-object** removes every object type; there is no per-type delete tool. Reminders use type \"reminder\", location reminders \"location_reminder\".\n- A workspace project must be archived with **project-management** before it can be deleted. A personal project can be deleted directly.\n\n**Templates**\n\n- Templates cannot be listed through this server, so only use an ID or URL the user gave you. To start a new project from one, call **add-projects** first and import into it. Imports write immediately and cannot be undone.\n\n**Project health**\n\n- Health data may be stale \u2014 check the isStale flag, and call **analyze-project-health** to refresh it before reading again.\n\n**Batches**\n\n- Prefer a batch tool over one call per item.\n- Where a batch tool returns per-item `failures` alongside successes, one failure does not undo the rest: never retry the whole batch, re-send only the items whose failure reason is fixable. **reschedule-tasks** is the exception \u2014 it has no per-item failures and throws if any task fails, so retry it as a whole.\n\nUse the **productivity-analysis** prompt for a combined productivity review \u2014 it pulls together user-info, get-productivity-stats and find-completed-tasks.\n\nAlways provide clear, actionable task titles and descriptions. Use the overview tools to give users context about their workload and project status.\n";
 /**
  * Create the MCP server.
  * @param todoistApiKey - The API key for the todoist account.

package/dist/mcp-server.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"mcp-server.d.ts","sourceRoot":"","sources":["../src/mcp-server.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,SAAS,EAAE,MAAM,yCAAyC,CAAA;AAGnE,OAAO,EACH,aAAa,EACb,KAAK,OAAO,EACZ,KAAK,WAAW,EAChB,KAAK,QAAQ,EAEhB,MAAM,kBAAkB,CAAA;AAKzB,eAAO,MAAM,YAAY,ksHAmDxB,CAAA;AAED;;;;;;GAMG;AACH,iBAAS,YAAY,CAAC,EAClB,aAAa,EACb,OAAO,EACP,QAAa,GAChB,EAAE;IACC,aAAa,EAAE,MAAM,CAAA;IACrB,OAAO,CAAC,EAAE,MAAM,CAAA;IAChB,QAAQ,CAAC,EAAE,QAAQ,CAAA;CACtB,aA4CA;AAED,OAAO,EAAE,aAAa,EAAE,KAAK,OAAO,EAAE,KAAK,WAAW,EAAE,KAAK,QAAQ,EAAE,YAAY,EAAE,CAAA"}
+{"version":3,"file":"mcp-server.d.ts","sourceRoot":"","sources":["../src/mcp-server.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,SAAS,EAAE,MAAM,yCAAyC,CAAA;AAGnE,OAAO,EACH,aAAa,EACb,KAAK,OAAO,EACZ,KAAK,WAAW,EAChB,KAAK,QAAQ,EAEhB,MAAM,kBAAkB,CAAA;AAKzB,eAAO,MAAM,YAAY,m1HAmDxB,CAAA;AAED;;;;;;GAMG;AACH,iBAAS,YAAY,CAAC,EAClB,aAAa,EACb,OAAO,EACP,QAAa,GAChB,EAAE;IACC,aAAa,EAAE,MAAM,CAAA;IACrB,OAAO,CAAC,EAAE,MAAM,CAAA;IAChB,QAAQ,CAAC,EAAE,QAAQ,CAAA;CACtB,aA4CA;AAED,OAAO,EAAE,aAAa,EAAE,KAAK,OAAO,EAAE,KAAK,WAAW,EAAE,KAAK,QAAQ,EAAE,YAAY,EAAE,CAAA"}

package/dist/{require-valid-todoist-token-DFE6h87D.js → require-valid-todoist-token-CC6hWyEx.js}

@@ -1,4 +1,4 @@
-import { Z as A, _, $ as T } from "./mcp-server-Cm72B6O_.js";
+import { Z as A, _, $ as T } from "./mcp-server-Drv9wBc-.js";
 const v = /* @__PURE__ */ new Set([401, 403]);
 function y(e) {
   const t = T(e);

package/dist/tools/find-activity.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"find-activity.d.ts","sourceRoot":"","sources":["../../src/tools/find-activity.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAA;AA6EvB,QAAA,MAAM,YAAY;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAsD6C,CAAA;AA6I/D,OAAO,EAAE,YAAY,EAAE,CAAA"}
+{"version":3,"file":"find-activity.d.ts","sourceRoot":"","sources":["../../src/tools/find-activity.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAA;AAkFvB,QAAA,MAAM,YAAY;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAsD6C,CAAA;AA6I/D,OAAO,EAAE,YAAY,EAAE,CAAA"}

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@doist/todoist-mcp",
-    "version": "12.5.5",
+    "version": "12.5.6",
     "type": "module",
     "main": "./dist/index.js",
     "types": "./dist/index.d.ts",

package/scripts/eval-instructions.ts

@@ -14,9 +14,11 @@
  *   git stash pop
  *   npx tsx scripts/eval-instructions.ts --label after
  *
- * Tool calls are never executed — only the first call of each turn is
- * inspected — so this touches no Todoist data and needs no Todoist token. It
- * does spend money on model calls; see --repeats and --models.
+ * Tool calls are never executed against Todoist, so this touches no Todoist
+ * data and needs no Todoist token. Judging looks at one call per attempt: the
+ * first, unless it is a context tool (see CONTEXT_TOOLS), in which case a
+ * canned result is fed back and the next call is judged. It does spend money
+ * on model calls; see --repeats and --models.
  *
  * Auth uses the Anthropic SDK's standard credential chain: either
  * ANTHROPIC_API_KEY (a key from the Anthropic Console at platform.claude.com),
@@ -34,9 +36,17 @@ import Anthropic from '@anthropic-ai/sdk'
 import { z } from 'zod'
 import { instructions } from '../src/mcp-server.js'
 import { registeredTools } from '../src/tool-registry.js'
+import type { UserInfoStructured } from '../src/tools/user-info.js'
 import { createLimiter } from '../src/utils/concurrency.js'
 import { ToolNames } from '../src/utils/tool-names.js'
 
+/**
+ * The user ID the stubbed `user-info` reports. A scenario asking a first-person
+ * question can assert the model carried it into a filter argument rather than
+ * querying everyone.
+ */
+const EVAL_USER_ID = '2671355'
+
 type Check = (input: Record<string, unknown>) => string | null
 
 /** First item of a batch tool's array argument, e.g. add-tasks' `tasks`. */
@@ -89,9 +99,14 @@ const SCENARIOS: Scenario[] = [
             if (!input.dateFrom || !input.dateTo) {
                 return 'no dateFrom/dateTo, so this searches all history rather than last week'
             }
+            // find-activity reports every user's events by default, so "what did
+            // I get done" answered without this includes collaborators' work.
+            if (input.initiatorId !== EVAL_USER_ID) {
+                return `initiatorId is ${JSON.stringify(input.initiatorId)}, so this reports every collaborator's completions, not "I"`
+            }
             return null
         },
-        guards: 'instructions: find-activity vs find-completed-tasks, over a bounded range',
+        guards: 'instructions: find-activity vs find-completed-tasks, over a bounded range, filtered to the asker',
     },
     {
         id: 'resolve-person',
@@ -195,6 +210,51 @@ const DEFAULT_MODELS = ['claude-haiku-4-5', 'claude-sonnet-5']
 const DEFAULT_REPEATS = 5
 const MAX_CONCURRENCY = 4
 
+/**
+ * Tools a model may legitimately call before the one a scenario is testing, to
+ * establish context it has no other way to get — today's date, the user's
+ * timezone. "What did I get done last week?" cannot be turned into a date range
+ * without them.
+ *
+ * Judging the very first call outright scores that correct behaviour as a
+ * failure: an `expect` allowlist naming the tool under test cannot also name
+ * every reasonable lookup that precedes it. So a call to one of these is
+ * answered with the canned result below and judging moves to the next call.
+ *
+ * The result is fiction, but its shape has to be real: a scenario may assert
+ * that the model carried a value through (see EVAL_USER_ID), and a model reads
+ * a malformed field the way it would read any other bad data. `satisfies` ties
+ * it to the tool's own output type, so a renamed or added field breaks here
+ * rather than drifting quietly. Formats must match too — `currentLocalTime`
+ * is what `toLocaleString('en-US', …)` produces, not ISO.
+ */
+const CONTEXT_TOOLS: Record<string, unknown> = {
+    [ToolNames.USER_INFO]: {
+        type: 'user_info',
+        userId: EVAL_USER_ID,
+        fullName: 'Eval User',
+        timezone: 'Europe/London',
+        currentLocalTime: '08/11/2026, 09:30:00',
+        startDay: 1,
+        startDayName: 'Monday',
+        weekStartDate: '2026-08-10',
+        weekEndDate: '2026-08-16',
+        currentWeekNumber: 33,
+        completedToday: 3,
+        dailyGoal: 5,
+        weeklyGoal: 30,
+        email: 'eval@example.com',
+        plan: 'Todoist Pro',
+    } satisfies UserInfoStructured,
+}
+
+/**
+ * How many context calls an attempt may make before it is judged a failure.
+ * A model that keeps gathering context is not answering the question, and
+ * without a cap a loop would bill for turns forever.
+ */
+const MAX_CONTEXT_HOPS = 3
+
 function parseArgs() {
     const args = process.argv.slice(2)
     const get = (flag: string) => {
@@ -254,6 +314,28 @@ type Attempt = {
      */
     errored: boolean
     usage: Usage
+    /** Context calls answered with a stub before the judged one. */
+    contextHops: number
+}
+
+/** Verdict on the one call a scenario is judged by. */
+function judge(
+    scenario: Scenario,
+    call: Anthropic.ToolUseBlock,
+): { pass: boolean; reason: string } {
+    if (scenario.forbid?.includes(call.name)) {
+        return { pass: false, reason: `called ${call.name}, which this rule forbids` }
+    }
+    if (scenario.expect && !scenario.expect.includes(call.name)) {
+        return { pass: false, reason: `expected ${scenario.expect.join(' or ')}` }
+    }
+    // An argument check written for a specific tool must not run against a
+    // different one a forbid-only scenario legitimately allows.
+    if (scenario.forbid && !scenario.expect) {
+        return { pass: true, reason: '' }
+    }
+    const reason = scenario.check?.(call.input as Record<string, unknown>) ?? null
+    return { pass: reason === null, reason: reason ?? '' }
 }
 
 async function runAttempt(
@@ -267,51 +349,77 @@ async function runAttempt(
         model,
         errored: false,
         usage: { input: 0, output: 0, cacheWrite: 0, cacheRead: 0 },
+        contextHops: 0,
     }
+    const messages: Anthropic.MessageParam[] = [{ role: 'user', content: scenario.prompt }]
     try {
-        const response = await client.messages.create({
-            model,
-            max_tokens: 4096,
-            // Tools render before system, so one breakpoint here caches both.
-            system: [{ type: 'text', text: instructions, cache_control: { type: 'ephemeral' } }],
-            tools,
-            messages: [{ role: 'user', content: scenario.prompt }],
-        })
+        for (let hop = 0; ; hop++) {
+            const response = await client.messages.create({
+                model,
+                max_tokens: 4096,
+                // Tools render before system, so one breakpoint here caches both.
+                system: [
+                    { type: 'text', text: instructions, cache_control: { type: 'ephemeral' } },
+                ],
+                tools,
+                messages,
+            })
 
-        base.usage = {
-            input: response.usage.input_tokens,
-            output: response.usage.output_tokens,
-            cacheWrite: response.usage.cache_creation_input_tokens ?? 0,
-            cacheRead: response.usage.cache_read_input_tokens ?? 0,
-        }
+            base.usage = {
+                input: base.usage.input + response.usage.input_tokens,
+                output: base.usage.output + response.usage.output_tokens,
+                cacheWrite:
+                    base.usage.cacheWrite + (response.usage.cache_creation_input_tokens ?? 0),
+                cacheRead: base.usage.cacheRead + (response.usage.cache_read_input_tokens ?? 0),
+            }
 
-        const call = response.content.find((b) => b.type === 'tool_use')
-        if (!call) {
-            return { ...base, calledTool: null, pass: false, reason: 'no tool call' }
-        }
-        if (scenario.forbid?.includes(call.name)) {
-            return {
-                ...base,
-                calledTool: call.name,
-                pass: false,
-                reason: `called ${call.name}, which this rule forbids`,
+            const calls = response.content.filter((b) => b.type === 'tool_use')
+            const call = calls[0]
+            if (!call) {
+                return { ...base, calledTool: null, pass: false, reason: 'no tool call' }
             }
-        }
-        if (scenario.expect && !scenario.expect.includes(call.name)) {
-            return {
-                ...base,
-                calledTool: call.name,
-                pass: false,
-                reason: `expected ${scenario.expect.join(' or ')}`,
+
+            // A turn can carry several calls at once. Judge the substantive one
+            // rather than whichever block came first: a model that asks for the
+            // date and queries the activity log in the same turn has made its
+            // choice, and the order between the two is arbitrary. A forbidden
+            // call outranks that — catching it is the point of a forbid rule.
+            //
+            // A tool a scenario expects is never treated as context, or a
+            // scenario testing the route *to* user-info could never pass: its
+            // expected call would be stubbed instead of judged.
+            const isContext = (c: Anthropic.ToolUseBlock) =>
+                Object.hasOwn(CONTEXT_TOOLS, c.name) && !scenario.expect?.includes(c.name)
+            const substantive =
+                calls.find((c) => scenario.forbid?.includes(c.name)) ??
+                calls.find((c) => !isContext(c))
+            if (substantive) {
+                const { pass, reason } = judge(scenario, substantive)
+                return { ...base, calledTool: substantive.name, pass, reason: reason || null }
             }
+
+            // Nothing but context gathering, so answer it and judge what the
+            // model reaches for next.
+            if (hop >= MAX_CONTEXT_HOPS) {
+                return {
+                    ...base,
+                    calledTool: call.name,
+                    pass: false,
+                    reason: `still gathering context after ${MAX_CONTEXT_HOPS} calls`,
+                }
+            }
+
+            base.contextHops = hop + 1
+            messages.push({ role: 'assistant', content: response.content })
+            messages.push({
+                role: 'user',
+                content: calls.map((c) => ({
+                    type: 'tool_result' as const,
+                    tool_use_id: c.id,
+                    content: JSON.stringify(CONTEXT_TOOLS[c.name]),
+                })),
+            })
         }
-        // An argument check written for a specific tool must not run against a
-        // different one a forbid-only scenario legitimately allows.
-        if (scenario.forbid && !scenario.expect) {
-            return { ...base, calledTool: call.name, pass: true, reason: null }
-        }
-        const reason = scenario.check?.(call.input as Record<string, unknown>) ?? null
-        return { ...base, calledTool: call.name, pass: reason === null, reason }
     } catch (error) {
         const message = error instanceof Error ? error.message : String(error)
         return {