@moda-ai/cli 1.48.0 → 1.50.0

package/dist/{cli-fpt3jkyh.js → cli-7kbzqact.js}

@@ -7,7 +7,7 @@ import {
   saveProfileConfig,
   secretFilePathFromRef,
   writeSecretJsonFile
-} from "./cli-233xm35b.js";
+} from "./cli-xrs02c0t.js";
 import {
   __require
 } from "./cli-0v6na3yp.js";

package/dist/{cli-ss2nere4.js → cli-h4k23fb0.js}

@@ -1,7 +1,7 @@
 import {
   authFetch,
   getCachedLoginContext
-} from "./cli-fpt3jkyh.js";
+} from "./cli-7kbzqact.js";
 import {
   CliAccessError,
   CliInputError,
@@ -15,7 +15,7 @@ import {
   resolveTenantIdFromActiveProfile,
   saveCliConfig,
   writeSecretJsonFile
-} from "./cli-233xm35b.js";
+} from "./cli-xrs02c0t.js";
 
 // src/init/tenant.ts
 import { randomUUID } from "node:crypto";

package/dist/{cli-eaaegm7a.js → cli-jqfm8dp4.js}

@@ -13,7 +13,7 @@ import {
   resolvePromptSource,
   safePromptEvidencePath,
   scanHarness
-} from "./cli-tznhpmf2.js";
+} from "./cli-x0x0n35w.js";
 import {
   SKILL_SOURCE_HARNESS,
   SKILL_SOURCE_INTEGRATION_BUNDLED,
@@ -26,7 +26,7 @@ import {
 import {
   isAuthSessionValid,
   loadAuthSession
-} from "./cli-fpt3jkyh.js";
+} from "./cli-7kbzqact.js";
 import {
   CliInputError,
   EXIT_ERROR,
@@ -45,7 +45,7 @@ import {
   resolveTenantId,
   secretFilePathFromRef,
   validateConfig
-} from "./cli-233xm35b.js";
+} from "./cli-xrs02c0t.js";
 import {
   __commonJS,
   __require,
@@ -6711,7 +6711,7 @@ async function runPromptSyncExclusive(flags, profileOptions, options) {
   const analyzedDefinitions = new Map;
   if (flags["no-analyze"] !== "true") {
     options.onProgress?.("Analyzing prompt sources before sync.");
-    const { analyzeHarnessPrompts } = await import("./harness-vf1e1t2h.js");
+    const { analyzeHarnessPrompts } = await import("./harness-7gr0xbtc.js");
     const result = await analyzeHarnessPrompts({
       rootDir: process.cwd(),
       existing: established.flatMap((prompt) => prompt.sourceDefinition ? [prompt.sourceDefinition.source] : []),

package/dist/{cli-tznhpmf2.js → cli-x0x0n35w.js}

@@ -25,7 +25,7 @@ import {
   resolveTenantId,
   shouldInventoryFile,
   stringOption
-} from "./cli-233xm35b.js";
+} from "./cli-xrs02c0t.js";
 import {
   __commonJS,
   __require,
@@ -177909,13 +177909,59 @@ var apiRequestCount = 0;
 class ApiError extends Error {
   statusCode;
   responseBody;
+  code;
+  suggestions;
+  readable;
   constructor(message, statusCode, responseBody) {
-    super(message);
+    const typed = parseTypedErrorBody(responseBody);
+    const platform = platformFailure(responseBody);
+    const readable = typed?.code === "audit_too_large" || platform !== null;
+    super(readable ? `API error (${statusCode}): ${typed?.message ?? platform?.message}` : message);
     this.statusCode = statusCode;
     this.responseBody = responseBody;
     this.name = "ApiError";
+    this.code = typed?.code ?? platform?.code;
+    this.suggestions = typed?.suggestions;
+    this.readable = readable;
+  }
+}
+function parseTypedErrorBody(body) {
+  if (!body.trimStart().startsWith("{"))
+    return null;
+  try {
+    const parsed = JSON.parse(body);
+    if (typeof parsed.code !== "string")
+      return null;
+    const suggestions = Array.isArray(parsed.suggestions) ? parsed.suggestions.filter((s) => typeof s === "string") : undefined;
+    return {
+      code: parsed.code,
+      message: typeof parsed.message === "string" ? parsed.message : undefined,
+      suggestions
+    };
+  } catch {
+    return null;
   }
 }
+var DETERMINISTIC_PLATFORM_FAILURES = [
+  "FUNCTION_INVOCATION_FAILED",
+  "FUNCTION_RESPONSE_PAYLOAD_TOO_LARGE"
+];
+function platformFailure(body) {
+  const code = DETERMINISTIC_PLATFORM_FAILURES.find((c) => body.includes(c));
+  if (!code)
+    return null;
+  const requestId = /\b([a-z]{3}\d::[A-Za-z0-9:-]+)/.exec(body)?.[1];
+  const cause = code === "FUNCTION_RESPONSE_PAYLOAD_TOO_LARGE" ? "the response was larger than the hosting platform allows" : "the server function crashed (typically out of memory or an oversized response)";
+  return {
+    code,
+    message: `The Moda API could not complete this request: ${cause} [${code}` + `${requestId ? `, request ${requestId}` : ""}]. Retrying the same request will fail the same way; ` + "narrow it (a smaller --limit, a single trace, or fewer results) and report it with `moda feedback` if it persists."
+  };
+}
+function isRetryableResponse(status, body) {
+  if (!isRetryable(status))
+    return false;
+  return platformFailure(body) === null;
+}
 
 class NetworkError extends Error {
   constructor(message) {
@@ -177970,7 +178016,7 @@ async function callModaAPI(url, options, profileOptions = {}, requestOptions = {
       if (!response.ok) {
         const errorText = await response.text();
         const error = new ApiError(`API error (${response.status}): ${errorText}`, response.status, errorText);
-        if (isRetryable(response.status) && attempt < maxRetries) {
+        if (isRetryableResponse(response.status, errorText) && attempt < maxRetries) {
           continue;
         }
         throw error;
@@ -180164,7 +180210,7 @@ function citationExcerpt(rootDir, path, line) {
     const selected = line ? lines[line - 1] : lines.find((candidate) => candidate.trim().length > 0);
     if (!selected)
       return {};
-    return { excerpt: selected.trim().slice(0, 240) };
+    return { excerpt: selected.trim() };
   } catch {
     return {};
   }
@@ -181019,7 +181065,6 @@ function enrichScanCaptureMetadata(graph) {
 // src/harness-candidates.ts
 import { mkdirSync as mkdirSync3, writeFileSync as writeFileSync3 } from "node:fs";
 import { basename as basename3, dirname as dirname3, extname, resolve as resolve4 } from "node:path";
-var EXCERPT_MAX_CHARS = 240;
 var MAX_CANDIDATES_PER_FILE_PER_SIGNAL = 20;
 var MIN_PROMPT_LITERAL_CHARS = 160;
 var CONFIDENCE_WEIGHT = {
@@ -181265,8 +181310,7 @@ function instructionScore(text) {
   return score;
 }
 function normalizeExcerpt(text) {
-  const collapsed = text.replace(/\s+/g, " ").trim();
-  return collapsed.length > EXCERPT_MAX_CHARS ? `${collapsed.slice(0, EXCERPT_MAX_CHARS - 1)}…` : collapsed;
+  return text.replace(/\s+/g, " ").trim();
 }
 function buildLineIndex(text) {
   const offsets = [0];
@@ -181377,7 +181421,7 @@ function detectPromptModule(file, out) {
 `).length),
     confidence: "medium",
     signal: "prompt_module.name_hint",
-    excerpt: normalizeExcerpt(file.text.slice(0, EXCERPT_MAX_CHARS * 2))
+    excerpt: normalizeExcerpt(file.text)
   });
 }
 function detectPromptLiterals(file, out) {
@@ -181460,7 +181504,7 @@ function detectPromptFile(file, out) {
 `).length),
     confidence,
     signal,
-    excerpt: normalizeExcerpt(file.text.slice(0, EXCERPT_MAX_CHARS * 2))
+    excerpt: normalizeExcerpt(file.text)
   });
 }
 function detectPromptLoaderReferences(file, filesByPath, out) {
@@ -184813,7 +184857,7 @@ async function runHarnessCommand(context) {
         throw new CliInputError("Focused remote prompt analysis is not supported yet.", "Use local --analyst=claude|codex|cursor|local-scan, or drop --concern for a full remote report.");
       }
       if (context.flags["github-actions"] === "true") {
-        const { runGithubActionsAnalyze } = await import("./harness-github-actions-379faaqe.js");
+        const { runGithubActionsAnalyze } = await import("./harness-github-actions-q9kewzb6.js");
         const result = await runGithubActionsAnalyze(rootDir, {
           writeReport: (report2) => {
             const normalized = normalizeHarnessReport(report2);
@@ -186559,6 +186603,7 @@ var CONCERN_CANDIDATE_KINDS = {
   evals: ["model_config"]
 };
 var MAX_SEED_CANDIDATES_PER_KIND = 40;
+var MAX_INLINE_CANDIDATE_EXCERPT_CHARS = 500;
 var MAX_SEED_READ_PLAN_FILES = 25;
 var SEED_CANDIDATE_KIND_LABELS = {
   llm_call_site: "LLM call sites",
@@ -186598,7 +186643,7 @@ function renderAnalystCandidateLines(report, concern) {
     const shown = bucket.slice(0, MAX_SEED_CANDIDATES_PER_KIND);
     lines.push(`${label} (${bucket.length}${bucket.length > shown.length ? `, showing ${shown.length}` : ""}):`);
     for (const candidate of shown) {
-      lines.push(`- ${candidate.path}:${candidate.startLine} [${candidate.signal}] ${candidate.excerpt}`);
+      lines.push(candidate.endLine > candidate.startLine ? `- ${candidate.path}:${candidate.startLine}-${candidate.endLine} [${candidate.signal}] (multi-line span; read lines ${candidate.startLine}-${candidate.endLine} of the file)` : candidate.excerpt.length > MAX_INLINE_CANDIDATE_EXCERPT_CHARS ? `- ${candidate.path}:${candidate.startLine} [${candidate.signal}] (${candidate.excerpt.length}-character match; read line ${candidate.startLine} of the file)` : `- ${candidate.path}:${candidate.startLine} [${candidate.signal}] ${candidate.excerpt}`);
     }
   }
   const readPlan = report.files.slice(0, MAX_SEED_READ_PLAN_FILES);

package/dist/{cli-233xm35b.js → cli-xrs02c0t.js}

@@ -986,6 +986,9 @@ function packSnapshot(inputs, opts) {
 }
 // src/output.ts
 import { randomUUID } from "node:crypto";
+import { lstatSync, mkdirSync as mkdirSync2, readdirSync, unlinkSync, writeFileSync as writeFileSync2 } from "node:fs";
+import { tmpdir } from "node:os";
+import { join as join2 } from "node:path";
 var AGENT_SCHEMA_VERSION = "moda.agent.v1";
 var AGENT_EVENT_SCHEMA_VERSION = "moda.agent_event.v1";
 var FEEDBACK_TIP = 'Help us improve! If a result in this response looks wrong or is missing data, flag it: moda feedback "<describe what looked wrong>" [--category=...] [--conversation-id=... | --cluster-id=... | --tool-name=...]. Only submit genuine observations.';
@@ -1005,7 +1008,8 @@ var AGENT_ERROR_CODES = [
   "command_error",
   "access_denied",
   "unknown_error",
-  "cloud_unavailable"
+  "cloud_unavailable",
+  "output_error"
 ];
 var FALLBACK_RUN_ID = randomUUID();
 var fallbackSeq = 0;
@@ -1116,6 +1120,7 @@ function createCliOutput(options) {
     seq: 0,
     started: false,
     terminal: false,
+    outputError: false,
     startedAt: Date.now(),
     lastEventAt: Date.now()
   };
@@ -1177,6 +1182,8 @@ function createCliOutput(options) {
         if (state.terminal)
           return;
         const envelope = writeEnvelope(data, writeOptions);
+        if (envelope.errors.some((error) => error.code === "output_error"))
+          state.outputError = true;
         if (options.mode === "agent-stream") {
           ensureStarted();
           emitEvent({
@@ -1272,6 +1279,9 @@ function createCliOutput(options) {
     hasTerminalOutput() {
       return state.terminal;
     },
+    hasOutputError() {
+      return state.outputError;
+    },
     close() {
       clearHeartbeat(state);
     }
@@ -1281,41 +1291,198 @@ function createCliOutput(options) {
 function createAgentEnvelope(options) {
   const safeData = redactDeep(options.data);
   const maxBytes = agentMaxOutputBytes();
+  let fullOutput;
+  let saveFailed = false;
+  let envelopeTextSaveFailed = false;
+  const writeFullOutput = options.persistFullOutput === false ? undefined : () => {
+    fullOutput ??= { path: writeAgentFullOutput(options.command, safeData) };
+    if (!fullOutput.path)
+      saveFailed = true;
+    return fullOutput.path;
+  };
+  let text = redactDeep({
+    summary: normalizeSummary(options.summary, options.command, options.status, safeData),
+    findings: options.findings ?? [],
+    evidence_refs: options.evidenceRefs ?? [],
+    recommended_actions: options.recommendedActions ?? [],
+    next_commands: normalizeNextCommands(options.nextCommands ?? []),
+    warnings: options.warnings ?? inferWarnings(safeData)
+  });
+  let moved;
   const build = (bounded) => ({
     schema_version: AGENT_SCHEMA_VERSION,
     command: options.command,
     status: options.status,
     generated_at: new Date().toISOString(),
     ...options.profile ? { profile: options.profile } : {},
-    summary: normalizeSummary(options.summary, options.command, options.status, safeData),
-    findings: options.findings ?? [],
-    evidence_refs: options.evidenceRefs ?? [],
-    recommended_actions: options.recommendedActions ?? [],
-    next_commands: normalizeNextCommands(options.nextCommands ?? []),
-    artifacts: options.artifacts ?? [],
+    summary: text.summary,
+    findings: text.findings,
+    evidence_refs: text.evidence_refs,
+    recommended_actions: text.recommended_actions,
+    next_commands: text.next_commands,
+    artifacts: [
+      ...options.artifacts ?? [],
+      ...bounded.fullOutputPath ? [{ kind: "full_output", path: bounded.fullOutputPath, description: "Complete command output (the same JSON --json prints); data in this envelope was bounded." }] : [],
+      ...moved?.path ? [{ kind: "full_envelope_text", path: moved.path, description: "Whole envelope text (summary, findings, evidence_refs, actions, next_commands, warnings) before oversized fields were moved out of this envelope." }] : []
+    ],
     warnings: [
-      ...options.warnings ?? inferWarnings(safeData),
-      ...bounded.warning ? [bounded.warning] : []
+      ...text.warnings,
+      ...bounded.warning ? [bounded.warning] : [],
+      ...moved ? [movedTextWarning(moved)] : []
     ],
     errors: options.errors ?? [],
     data: bounded.data,
     ...options.meta ? { meta: options.meta } : {},
-    ...bounded.truncation ? { truncation: bounded.truncation } : {}
+    ...bounded.truncation || moved ? { truncation: {
+      ...bounded.truncation ?? {},
+      ...moved ? { envelope_text: {
+        reason: "agent_output_byte_limit",
+        moved_fields: moved.fields,
+        complete: false,
+        ...moved.whole ? { whole_text_moved: true } : {},
+        ...moved.path ? { full_envelope_path: moved.path } : {}
+      } } : {}
+    } } : {}
   });
-  let envelope = build(boundForAgent(safeData, maxBytes));
-  const envelopeBytes = Buffer.byteLength(JSON.stringify(envelope), "utf8");
-  if (envelopeBytes > maxBytes) {
-    const overhead = envelopeBytes - Buffer.byteLength(JSON.stringify(envelope.data) ?? "", "utf8");
-    envelope = build(boundForAgent(safeData, Math.max(1024, maxBytes - overhead - 1024)));
+  const sizeOf = (value) => Buffer.byteLength(JSON.stringify(value) ?? "", "utf8");
+  const rebound = (current) => {
+    const overhead = sizeOf(current) - sizeOf(current.data);
+    return build(boundForAgent(safeData, Math.max(1024, maxBytes - overhead - 1024), { writeFullOutput }));
+  };
+  let envelope = build(boundForAgent(safeData, maxBytes, { writeFullOutput }));
+  if (sizeOf(envelope) > maxBytes)
+    envelope = rebound(envelope);
+  if (sizeOf(envelope) > maxBytes && !saveFailed) {
+    const original = text;
+    const savedPath = options.persistFullOutput === false ? undefined : writeAgentFullOutput(`${options.command}-envelope`, original);
+    if (options.persistFullOutput !== false && !savedPath) {
+      saveFailed = true;
+      envelopeTextSaveFailed = true;
+    } else {
+      const slots = envelopeTextSlots(original).sort((left, right) => right.bytes - left.bytes);
+      text = structuredClone(original);
+      moved = { path: savedPath, fields: [] };
+      for (const slot of slots) {
+        if (sizeOf(envelope) <= maxBytes)
+          break;
+        const pointer = movedTextPointer(slot.field, slot.text, savedPath);
+        const added = Buffer.byteLength(JSON.stringify(pointer), "utf8") + 2 * movedFieldNameBytes(slot.field);
+        if (added >= slot.bytes)
+          continue;
+        slot.set(text, pointer);
+        moved.fields.push(slot.field);
+        envelope = rebound(envelope);
+      }
+      if (moved.fields.length === 0)
+        moved = undefined;
+      envelope = rebound(envelope);
+      if (sizeOf(envelope) > maxBytes) {
+        text = {
+          summary: { text: movedWholeTextPointer(savedPath), confidence: original.summary.confidence },
+          findings: [],
+          evidence_refs: [],
+          recommended_actions: [],
+          next_commands: [],
+          warnings: []
+        };
+        moved = { path: savedPath, fields: ["*"], whole: true };
+        envelope = rebound(envelope);
+      }
+    }
+  }
+  if (saveFailed) {
+    envelope = {
+      ...envelope,
+      status: "error",
+      errors: [...envelope.errors, saveFailedError(maxBytes, envelopeTextSaveFailed, fullOutput?.path)]
+    };
   }
   return redactDeep(envelope);
 }
+function saveFailedError(maxBytes, envelopeTextSaveFailed, dataPath) {
+  const where = "(MODA_AGENT_OUTPUT_DIR is unwritable or the disk is full)";
+  const remedy = "Point MODA_AGENT_OUTPUT_DIR at a writable directory, rerun with --json (no byte limit), or narrow the query.";
+  if (!envelopeTextSaveFailed) {
+    return {
+      code: "output_error",
+      cause: "full_output_save_failed",
+      message: `Output is over the ${maxBytes}-byte agent limit and the complete output could not be saved to a file ${where}, ` + `so nothing was moved out: this envelope holds the whole result, over the limit. ${remedy}`,
+      blocking: false
+    };
+  }
+  const dataState = dataPath ? `data is NOT the whole result: it was bounded to fit, and the complete data was saved to ${dataPath} ` + "(truncation.data.full_output_path); read it in line ranges or with jq. " : "data is the whole result. ";
+  return {
+    code: "output_error",
+    cause: "envelope_text_save_failed",
+    message: `Output is over the ${maxBytes}-byte agent limit and the oversized envelope text (summary, findings, evidence_refs, ` + `recommended_actions, next_commands, warnings) could not be saved to a file ${where}, so that text was kept whole ` + `in this envelope, over the limit. ${dataState}${remedy}`,
+    blocking: false
+  };
+}
+function envelopeTextSlots(text) {
+  const slots = [
+    { field: "summary.text", text: text.summary.text, set: (target, value) => {
+      target.summary.text = value;
+    } }
+  ];
+  text.findings.forEach((finding, index) => {
+    slots.push({ field: `findings[${index}].title`, text: finding.title, set: (target, value) => {
+      target.findings[index].title = value;
+    } });
+    if (finding.description) {
+      slots.push({ field: `findings[${index}].description`, text: finding.description, set: (target, value) => {
+        target.findings[index].description = value;
+      } });
+    }
+  });
+  text.evidence_refs.forEach((ref, index) => {
+    slots.push({ field: `evidence_refs[${index}].label`, text: ref.label, set: (target, value) => {
+      target.evidence_refs[index].label = value;
+    } });
+  });
+  text.recommended_actions.forEach((action, index) => {
+    slots.push({ field: `recommended_actions[${index}].title`, text: action.title, set: (target, value) => {
+      target.recommended_actions[index].title = value;
+    } });
+    if (action.rationale) {
+      slots.push({ field: `recommended_actions[${index}].rationale`, text: action.rationale, set: (target, value) => {
+        target.recommended_actions[index].rationale = value;
+      } });
+    }
+  });
+  text.next_commands.forEach((command, index) => {
+    slots.push({ field: `next_commands[${index}].purpose`, text: command.purpose, set: (target, value) => {
+      target.next_commands[index].purpose = value;
+    } });
+  });
+  text.warnings.forEach((warning, index) => {
+    slots.push({ field: `warnings[${index}]`, text: warning, set: (target, value) => {
+      target.warnings[index] = value;
+    } });
+  });
+  return slots.map((slot) => ({ ...slot, bytes: Buffer.byteLength(JSON.stringify(slot.text), "utf8") }));
+}
+function movedTextPointer(field, value, path) {
+  const chars = Array.from(value).length;
+  return path ? `[${chars} characters moved whole to ${path} at ${field}]` : `[${chars} characters over the agent byte limit; not saved. Rerun with --json or raise MODA_AGENT_MAX_OUTPUT_BYTES]`;
+}
+function movedFieldNameBytes(field) {
+  return Buffer.byteLength(JSON.stringify(field), "utf8") + 2;
+}
+function movedWholeTextPointer(path) {
+  return path ? `[Envelope text (summary, findings, evidence_refs, recommended_actions, next_commands, warnings) was over the agent byte limit and moved whole to ${path}]` : "[Envelope text was over the agent byte limit; not saved. Rerun with --json or raise MODA_AGENT_MAX_OUTPUT_BYTES]";
+}
+function movedTextWarning(moved) {
+  if (moved.whole) {
+    return "Envelope text outside data was over the agent byte limit even after moving its largest fields: all of it was moved whole " + (moved.path ? `to ${moved.path} (read it with jq)` : "(the file could not be saved; rerun with --json or raise MODA_AGENT_MAX_OUTPUT_BYTES)") + ".";
+  }
+  return `Envelope text outside data was over the agent byte limit: ${moved.fields.length} field(s) were moved whole ` + (moved.path ? `to ${moved.path} (same field paths; read it with jq)` : "(the file could not be saved; rerun with --json or raise MODA_AGENT_MAX_OUTPUT_BYTES)") + ` and replaced by pointers: ${moved.fields.join(", ")}.`;
+}
 var AGENT_ENVELOPE_RESERVE_BYTES = 2048;
 function agentEnvelopeFits(command, data, options = {}) {
   const maxBytes = agentMaxOutputBytes() - AGENT_ENVELOPE_RESERVE_BYTES;
   if (Buffer.byteLength(JSON.stringify(data) ?? "", "utf8") > maxBytes)
     return false;
-  const envelope = createAgentEnvelope({ command, status: "ok", data, ...options });
+  const envelope = createAgentEnvelope({ command, status: "ok", data, ...options, persistFullOutput: false });
   return !envelope.truncation && Buffer.byteLength(JSON.stringify(envelope), "utf8") <= maxBytes;
 }
 function errorWarnings(error) {
@@ -1418,10 +1585,14 @@ function errorToAgentError(error) {
     }
     if (isApiLikeError(error)) {
       const statusCode = "statusCode" in error ? error.statusCode : undefined;
+      const typedCode = "code" in error && typeof error.code === "string" ? error.code : undefined;
+      const suggestions = "suggestions" in error && Array.isArray(error.suggestions) ? error.suggestions.filter((s) => typeof s === "string") : [];
+      const http = typeof statusCode === "number" ? `HTTP ${statusCode}` : undefined;
       return {
         code: "api_error",
         message: error.message,
-        cause: typeof statusCode === "number" ? `HTTP ${statusCode}` : undefined,
+        cause: typedCode ? http ? `${typedCode} (${http})` : typedCode : http,
+        ...suggestions.length > 0 ? { suggested_commands: suggestions } : {},
         blocking: true
       };
     }
@@ -1555,41 +1726,61 @@ function inferWarnings(data) {
   }
   return [];
 }
-var AGENT_MAX_STRING_CHARS = 2000;
-var PROTECTED_STRING_KEY = /(cursor|token|_id$|^id$|url$|href$|^key$|hash$)/i;
 function agentMaxOutputBytes() {
   return Number(process.env.MODA_AGENT_MAX_OUTPUT_BYTES ?? 200000);
 }
-function boundForAgent(data, maxBytes = agentMaxOutputBytes()) {
+var AGENT_FULL_OUTPUT_FILE = /-\d{4}-\d{2}-\d{2}T\d{2}-\d{2}-\d{2}-\d{3}Z-[0-9a-f]{8}\.json$/;
+function agentOutputRetentionMs() {
+  const hours = Number(process.env.MODA_AGENT_OUTPUT_RETENTION_HOURS ?? 24);
+  return (Number.isFinite(hours) && hours > 0 ? hours : 24) * 3600000;
+}
+function pruneAgentFullOutputs(dir, keep, now = Date.now()) {
+  const cutoff = now - agentOutputRetentionMs();
+  let names;
+  try {
+    names = readdirSync(dir);
+  } catch {
+    return;
+  }
+  for (const name of names) {
+    if (!AGENT_FULL_OUTPUT_FILE.test(name))
+      continue;
+    const path = join2(dir, name);
+    if (path === keep)
+      continue;
+    try {
+      const stat = lstatSync(path);
+      if (stat.isFile() && stat.mtimeMs < cutoff)
+        unlinkSync(path);
+    } catch {}
+  }
+}
+function writeAgentFullOutput(command, data) {
+  try {
+    const dir = process.env.MODA_AGENT_OUTPUT_DIR || join2(tmpdir(), "moda-agent-output");
+    mkdirSync2(dir, { recursive: true, mode: 448 });
+    const slug = command.replace(/[^a-zA-Z0-9_-]+/g, "-").replace(/^-+|-+$/g, "") || "output";
+    const stamp = new Date().toISOString().replace(/[:.]/g, "-");
+    const path = join2(dir, `${slug}-${stamp}-${randomUUID().slice(0, 8)}.json`);
+    writeFileSync2(path, `${JSON.stringify(data, null, 2)}
+`, { encoding: "utf8", mode: 384 });
+    pruneAgentFullOutputs(dir, path);
+    return path;
+  } catch {
+    return;
+  }
+}
+function boundForAgent(data, maxBytes = agentMaxOutputBytes(), options = {}) {
   const sizeOf = (value) => Buffer.byteLength(JSON.stringify(value) ?? "", "utf8");
   const serialized = JSON.stringify(data) ?? "";
   const originalBytes = Buffer.byteLength(serialized, "utf8");
   if (originalBytes <= maxBytes)
     return { data };
-  let shortenedStrings = 0;
-  const shorten = (value, limit, key) => {
-    if (typeof value === "string") {
-      if (value.length <= limit || key !== undefined && PROTECTED_STRING_KEY.test(key))
-        return value;
-      shortenedStrings += 1;
-      return `${value.slice(0, limit)}… [truncated ${value.length - limit} chars]`;
-    }
-    if (Array.isArray(value))
-      return value.map((item) => shorten(item, limit));
-    if (value && typeof value === "object") {
-      return Object.fromEntries(Object.entries(value).map(([entryKey, item]) => [entryKey, shorten(item, limit, entryKey)]));
-    }
-    return value;
-  };
-  let bounded = data;
+  const fullOutputPath = options.writeFullOutput?.();
+  if (options.writeFullOutput && !fullOutputPath)
+    return { data, saveFailed: true };
+  const bounded = JSON.parse(serialized);
   let boundedBytes = originalBytes;
-  for (const limit of [AGENT_MAX_STRING_CHARS, 500, 200]) {
-    shortenedStrings = 0;
-    bounded = shorten(data, limit);
-    boundedBytes = sizeOf(bounded);
-    if (boundedBytes <= maxBytes)
-      break;
-  }
   const droppedItems = {};
   const collectArrays = (value, path, out) => {
     if (Array.isArray(value)) {
@@ -1626,39 +1817,35 @@ function boundForAgent(data, maxBytes = agentMaxOutputBytes()) {
     boundedBytes = sizeOf(bounded);
   }
   const dropped = Object.entries(droppedItems);
-  const parts = [
-    shortenedStrings > 0 ? `shortened ${shortenedStrings} long text field(s)` : "",
-    ...dropped.map(([path, count]) => `dropped the last ${count} item(s) of data.${path}`)
-  ].filter(Boolean);
-  const howToGetMore = "For complete results narrow the query (a smaller --limit, with --offset or --cursor paging where the command supports it), " + "rerun with --json (no byte limit), or raise MODA_AGENT_MAX_OUTPUT_BYTES.";
+  const whereIsTheRest = fullOutputPath ? `The complete output (every item, no text shortened) is in ${fullOutputPath}; read it in line ranges or query it with jq. ` : "The complete output could not be saved to a file. ";
+  const howToGetMore = whereIsTheRest + "Alternatively narrow the query (a smaller --limit, with --offset or --cursor paging where the command supports it), " + "rerun with --json (no byte limit), or raise MODA_AGENT_MAX_OUTPUT_BYTES.";
+  const pathField = fullOutputPath ? { full_output_path: fullOutputPath } : {};
   if (boundedBytes > maxBytes) {
-    const stub = { truncated: true, original_bytes: originalBytes, max_bytes: maxBytes, preview: serialized.slice(0, Math.max(0, maxBytes - 256)) };
-    for (let stubBytes = sizeOf(stub);stubBytes > maxBytes && stub.preview.length > 0; stubBytes = sizeOf(stub)) {
-      stub.preview = stub.preview.slice(0, Math.max(0, stub.preview.length - (stubBytes - maxBytes) - 16));
-    }
     return {
-      data: stub,
-      warning: `Output was ${originalBytes} bytes, over the ${maxBytes}-byte agent limit, and could not be reduced while keeping its structure; data is a string preview. ${howToGetMore}`,
+      data: { truncated: true, original_bytes: originalBytes, max_bytes: maxBytes, ...pathField },
+      warning: `Output was ${originalBytes} bytes, over the ${maxBytes}-byte agent limit, and could not be reduced while keeping its structure; data is a pointer stub. ${howToGetMore}`,
       truncation: {
-        data: { original_bytes: originalBytes, max_bytes: maxBytes, reason: "agent_output_byte_limit", structure_preserved: false, complete: false }
-      }
+        data: { original_bytes: originalBytes, max_bytes: maxBytes, reason: "agent_output_byte_limit", structure_preserved: false, complete: false, ...pathField }
+      },
+      fullOutputPath
     };
   }
   return {
     data: bounded,
-    warning: `Output was ${originalBytes} bytes, over the ${maxBytes}-byte agent limit: ${parts.join("; ")}. ${howToGetMore}`,
+    warning: `Output was ${originalBytes} bytes, over the ${maxBytes}-byte agent limit: ${dropped.map(([path, count]) => `dropped the last ${count} item(s) of data.${path}`).join("; ")}. ${howToGetMore}`,
     truncation: {
       data: {
         original_bytes: originalBytes,
         bounded_bytes: boundedBytes,
         max_bytes: maxBytes,
         reason: "agent_output_byte_limit",
-        shortened_strings: shortenedStrings,
         dropped_items: droppedItems,
         structure_preserved: true,
-        complete: false
+        complete: false,
+        ...pathField
       }
-    }
+    },
+    fullOutputPath
   };
 }
 function asRecord(value) {
@@ -1717,4 +1904,4 @@ function createCommandContext(input, options = {}) {
   };
 }
 
-export { PROD_BASE_URL, DEFAULT_INGEST_URL, getModaPaths, getModaDir, isLoopbackBaseUrl, assertValidProfileName, describeBaseUrlSource, loadProfileConfig, saveProfileConfig, getActiveProfile, setActiveProfile, upsertProfile, clearStoredApiKeyCredentials, secretFilePathFromRef, resolveSecretReference, resolveApiKeyFromActiveProfile, resolveTenantIdFromActiveProfile, ensureSecretFilePermissions, writeSecretJsonFile, CliInputError, CliAccessError, CliInputRequiredError, CliAuthError, CliCancellationError, isNetworkLikeError, isApiLikeError, VERSION, MODA_BASE_URL, loadCliConfig, saveCliConfig, resolveApiKey, embeddedTenantIdFromApiKey, resolveIngestUrl2 as resolveIngestUrl, resolveTenantId, resolveApiTenantId, resolveModaBaseUrl, validateConfig, MAX_SCANNED_FILE_BYTES, DEFAULT_MAX_INVENTORY_FILES, DEFAULT_MAX_INVENTORY_TEXT_BYTES, MAX_RECORDED_SKIPS, SKIPPED_DIRS, isSkippedDir, isSkippedPathPrefix, isEvalLikePath, shouldInventoryFile, isTextFile, isSecretishPath, SECRET_KEYS, redactDeep, redactString, packSnapshot, EXIT_OK, EXIT_ERROR, EXIT_DEGRADED, EXIT_AUTH_REQUIRED, EXIT_INPUT_REQUIRED, EXIT_CANCELLED, AGENT_ERROR_CODES, wasOutputCancelled, safeWriteStdout, safeWriteStderr, explicitOutputMode, resolveOutputMode, createCliOutput, agentEnvelopeFits, stringOption, createCommandContext };
+export { PROD_BASE_URL, DEFAULT_INGEST_URL, getModaPaths, getModaDir, isLoopbackBaseUrl, assertValidProfileName, describeBaseUrlSource, loadProfileConfig, saveProfileConfig, getActiveProfile, setActiveProfile, upsertProfile, clearStoredApiKeyCredentials, secretFilePathFromRef, resolveSecretReference, resolveApiKeyFromActiveProfile, resolveTenantIdFromActiveProfile, ensureSecretFilePermissions, writeSecretJsonFile, CliInputError, CliAccessError, CliInputRequiredError, CliAuthError, CliCancellationError, isNetworkLikeError, isApiLikeError, VERSION, MODA_BASE_URL, loadCliConfig, saveCliConfig, resolveApiKey, embeddedTenantIdFromApiKey, resolveIngestUrl2 as resolveIngestUrl, resolveTenantId, resolveApiTenantId, resolveModaBaseUrl, validateConfig, MAX_SCANNED_FILE_BYTES, DEFAULT_MAX_INVENTORY_FILES, DEFAULT_MAX_INVENTORY_TEXT_BYTES, MAX_RECORDED_SKIPS, SKIPPED_DIRS, isSkippedDir, isSkippedPathPrefix, isEvalLikePath, shouldInventoryFile, isTextFile, isSecretishPath, SECRET_KEYS, redactDeep, redactString, packSnapshot, EXIT_OK, EXIT_ERROR, EXIT_DEGRADED, EXIT_AUTH_REQUIRED, EXIT_INPUT_REQUIRED, EXIT_CANCELLED, AGENT_ERROR_CODES, wasOutputCancelled, safeWriteStdout, safeWriteStderr, explicitOutputMode, resolveOutputMode, createCliOutput, agentEnvelopeFits, agentMaxOutputBytes, stringOption, createCommandContext };

package/dist/cli.js

@@ -8,7 +8,7 @@ import {
   runPromptsCommand,
   runSkillsCommand,
   runStatusCommand
-} from "./cli-eaaegm7a.js";
+} from "./cli-jqfm8dp4.js";
 import {
   ApiError,
   HARNESS_REPORT_APPROVAL_PATH,
@@ -37,7 +37,7 @@ import {
   terminalStyles,
   unknownSignalsSubcommandError,
   validateHarnessReport
-} from "./cli-tznhpmf2.js";
+} from "./cli-x0x0n35w.js";
 import"./cli-ssfq84k5.js";
 import {
   authFetch,
@@ -46,7 +46,7 @@ import {
   loadAuthSession,
   login,
   revokeCliSessionToken
-} from "./cli-fpt3jkyh.js";
+} from "./cli-7kbzqact.js";
 import {
   AGENT_ERROR_CODES,
   CliAccessError,
@@ -65,6 +65,7 @@ import {
   PROD_BASE_URL,
   VERSION,
   agentEnvelopeFits,
+  agentMaxOutputBytes,
   assertValidProfileName,
   clearStoredApiKeyCredentials,
   createCliOutput,
@@ -96,7 +97,7 @@ import {
   validateConfig,
   wasOutputCancelled,
   writeSecretJsonFile
-} from "./cli-233xm35b.js";
+} from "./cli-xrs02c0t.js";
 import {
   __require
 } from "./cli-0v6na3yp.js";
@@ -389,7 +390,9 @@ var AuditSchema = z.object({
   id: z.string().min(1),
   kind: z.enum(["auto", "trace", "conversation"]).default("auto").optional(),
   include_raw: boolFlag(),
-  limit: z.number().min(1).max(1000).default(200).optional()
+  include_bodies: boolFlag(),
+  limit: z.number().min(1).max(1000).default(200).optional(),
+  cursor: z.string().min(1).max(512).optional()
 });
 var TIME_RANGES = ["all", "1h", "3d", "7d", "24h", "30d", "90d"];
 var SEARCH_MODES = ["keyword", "semantic", "hybrid"];
@@ -544,8 +547,16 @@ function buildFrustrationAnchor(row) {
   if (firstQuote) {
     signal = asString(firstQuote.signal);
     const quote = asString(firstQuote.quote);
-    if (quote !== null)
-      quote_preview = quote.slice(0, 200);
+    if (quote !== null) {
+      let end = 0;
+      let points = 0;
+      for (const ch of quote) {
+        if (points++ === 200)
+          break;
+        end += ch.length;
+      }
+      quote_preview = quote.slice(0, end);
+    }
   }
   return {
     kind: "frustration",
@@ -621,7 +632,6 @@ function withFrustrationAnchors(response) {
   });
   return { ...response, frustrations };
 }
-var TOOL_FAILURE_TEXT_LIMIT = 300;
 var TOOL_FAILURE_CONTEXT_RADIUS = 2;
 function shellWord(value) {
   return /^[\w@%+=:,./-]+$/.test(value) ? value : `'${value.replace(/'/g, `'\\''`)}'`;
@@ -671,26 +681,6 @@ function trimToolFailureConversations(response, opts) {
   });
   return { ...response, examples };
 }
-function compactToolFailureExamples(response, maxChars = TOOL_FAILURE_TEXT_LIMIT) {
-  if (!Array.isArray(response.examples))
-    return { data: response, truncated: 0 };
-  let truncated = 0;
-  const cap = (value) => {
-    if (typeof value === "string") {
-      if (value.length <= maxChars)
-        return value;
-      truncated += 1;
-      return `${value.slice(0, maxChars)}… [truncated ${value.length - maxChars} chars]`;
-    }
-    if (Array.isArray(value))
-      return value.map(cap);
-    if (value && typeof value === "object") {
-      return Object.fromEntries(Object.entries(value).map(([key, child]) => [key, cap(child)]));
-    }
-    return value;
-  };
-  return { data: { ...response, examples: response.examples.map(cap) }, truncated };
-}
 function withToolFailureAnchors(response) {
   if (!Array.isArray(response.examples))
     return response;
@@ -5382,7 +5372,6 @@ var MODA_DIR = ".moda";
 var HARNESS_PATH = join4(MODA_DIR, "harness.json");
 var HARNESS_REPORT_MARKDOWN_PATH = join4(MODA_DIR, "HARNESS_REPORT.md");
 var HARNESS_SUMMARY_PATH = join4(MODA_DIR, "HARNESS.md");
-var MAX_AGENT_EVIDENCE_BYTES = 24000;
 async function runAgentCommand(context) {
   const subcommand2 = context.positionals[0] ?? "context";
   switch (subcommand2) {
@@ -5735,7 +5724,7 @@ function expandEvidenceRef(ref, loaded, cwd) {
         label: artifact.description ?? artifact.path,
         path: artifact.path
       },
-      value: readBoundedModaArtifact(cwd, artifact.path),
+      value: readModaArtifact(cwd, artifact.path),
       artifact
     };
   }
@@ -5766,11 +5755,11 @@ function readCitationExcerpt(cwd, citation) {
     return citation.excerpt;
   const text = readFileSync5(absolute, "utf8");
   if (citation.line && citation.line > 0) {
-    return text.split(/\r?\n/)[citation.line - 1]?.slice(0, 1000) ?? citation.excerpt;
+    return text.split(/\r?\n/)[citation.line - 1] ?? citation.excerpt;
   }
-  return text.slice(0, 1000);
+  return text;
 }
-function readBoundedModaArtifact(cwd, artifactPath) {
+function readModaArtifact(cwd, artifactPath) {
   if (!artifactPath.startsWith(".moda/")) {
     throw new CliInputError(`Artifact '${artifactPath}' is outside .moda and cannot be expanded directly.`, "Use a citation evidence ref for source-file evidence.");
   }
@@ -5778,12 +5767,10 @@ function readBoundedModaArtifact(cwd, artifactPath) {
   if (!absolute || !existsSync4(absolute))
     throw missingEvidence(`artifact:${artifactPath}`);
   const size = statSync2(absolute).size;
-  const text = readFileSync5(absolute, "utf8");
   return {
     path: artifactPath,
     bytes: size,
-    truncated: size > MAX_AGENT_EVIDENCE_BYTES,
-    text: text.slice(0, MAX_AGENT_EVIDENCE_BYTES)
+    text: readFileSync5(absolute, "utf8")
   };
 }
 function safeResolve(cwd, path) {
@@ -7279,7 +7266,7 @@ function toolRows(data, scopedTool) {
 function firstQuote(row) {
   const quotes = Array.isArray(row?.user_quotes) ? row.user_quotes : [];
   const first = asRecord(quotes[0]);
-  return asString(first?.quote)?.slice(0, 160) ?? undefined;
+  return asString(first?.quote) ?? undefined;
 }
 function likelyToolLocation(toolName, hints) {
   const normalized = normalizeKey(toolName);
@@ -7592,7 +7579,10 @@ async function finishRun(context, outcome) {
     context.output.close();
     if (wasOutputCancelled())
       return EXIT_CANCELLED;
-    return outcome.exitCode ?? EXIT_OK;
+    const exitCode = outcome.exitCode ?? EXIT_OK;
+    if (context.output.hasOutputError() && (exitCode === EXIT_OK || exitCode === EXIT_DEGRADED))
+      return EXIT_ERROR;
+    return exitCode;
   }
   if (outcome.kind === "cancelled") {
     context.output.writeCancelled(outcome.reason ?? "Command cancelled.");
@@ -7656,6 +7646,8 @@ function buildManifest(commands) {
       "MODA_AUTH_TIMEOUT_MS",
       "MODA_HEARTBEAT_MS",
       "MODA_AGENT_MAX_OUTPUT_BYTES",
+      "MODA_AGENT_OUTPUT_DIR",
+      "MODA_AGENT_OUTPUT_RETENTION_HOURS",
       "MODA_CLI_TELEMETRY",
       "MODA_CLI_UPDATE_CHECK",
       "MODA_CLI_TIPS"
@@ -7919,8 +7911,7 @@ Commands:
   hallucinations                   Grounding detections: contradicted/verified outputs with rule breakdown
   laziness                         Laziness detections per trace (repeated calls, premature handoffs, ...)
   tool-failures                    Get tool failure overview
-  tool-failure-detail <tool_name>  Get per-tool failure detail (--limit=N examples, 1-20, default 5;
-                                   long example text is cut to 300 chars)
+  tool-failure-detail <tool_name>  Get per-tool failure detail (--limit=N examples, 1-20, default 5)
   problems                         Rank cross-signal Problems by root cause (what to fix first)
   problems remainder               Signals no Problem explains yet (--family, --cursor)
   problem-impact                   Live problem/use-case impact (--node-id, --problem-id, --mode, --days-back)
@@ -8367,7 +8358,62 @@ function auditCapStatus(data, limit) {
   const rowsCapped = hasMore || !serverSaid && rows >= limit;
   if (!rowsCapped && !spansCapped)
     return null;
-  return { rowsCapped, spansCapped, hasMoreKnown: hasMore, rows, spans };
+  const reason = typeof summary.truncated_reason === "string" ? summary.truncated_reason : null;
+  const cursorValue = summary.next_cursor ?? record.next_cursor;
+  const nextCursor = typeof cursorValue === "string" && cursorValue ? cursorValue : null;
+  return { rowsCapped, spansCapped, hasMoreKnown: hasMore, rows, spans, reason, nextCursor };
+}
+var AUDIT_MIN_PAGE_BYTES = 32 * 1024;
+var AUDIT_AGENT_RESERVE_BYTES = 16 * 1024;
+async function fetchAudit(id, query2, maxBytes) {
+  const pageQuery = new URLSearchParams(query2);
+  if (maxBytes !== undefined)
+    pageQuery.set("max_bytes", maxBytes.toString());
+  const queryString = pageQuery.toString() ? `?${pageQuery.toString()}` : "";
+  return callDataAPI(`/audit/${encodeURIComponent(id)}${queryString}`);
+}
+function auditOutput(command, params, data) {
+  const capped = auditCapStatus(data, params.limit ?? AUDIT_DEFAULT_LIMIT);
+  const auditRecord = asRecord(data);
+  if (!capped || !auditRecord) {
+    return { data: auditRecord ? { ...auditRecord, truncated: false, has_more: false, next_cursor: null } : data, options: {} };
+  }
+  const limit = params.limit ?? AUDIT_DEFAULT_LIMIT;
+  const perTrace = "audit one OTLP trace at a time with `moda audit <trace_id> --kind=trace` (ids in summary.trace_ids)";
+  const warnings = [];
+  const nextCommands = [];
+  const auditFlags = (extra) => [
+    `moda ${command} ${shellArg(params.id)}`,
+    params.kind && params.kind !== "auto" ? `--kind=${params.kind}` : "",
+    params.include_raw === true ? "--include-raw" : "",
+    params.include_bodies === true ? "--include-bodies" : "",
+    ...extra
+  ].filter(Boolean).join(" ");
+  if (capped.nextCursor) {
+    nextCommands.push({
+      command: auditFlags([params.limit ? `--limit=${params.limit}` : "", `--cursor=${capped.nextCursor}`]),
+      purpose: "Read the next page of ingest rows of this trace.",
+      mutability: "read",
+      requires_approval: false
+    });
+  }
+  if (capped.reason === "byte_limit") {
+    warnings.push(`Spans were capped: the response reached the server's size budget after ${capped.rows} ingest row(s), so this trace has more spans than the ${capped.spans} shown. ` + `Continue with the next_cursor page${params.include_bodies === true ? ", or drop --include-bodies to fit more rows per page" : ""}. Orphan/duplicate analysis covers this page only.`);
+  } else if (capped.reason === "span_limit" && capped.nextCursor) {
+    warnings.push(`Spans were capped: the page reached the server's per-page span cap after ${capped.rows} ingest row(s), so this trace has more spans than the ${capped.spans} shown. ` + "Continue with the next_cursor page. Orphan/duplicate analysis covers this page only.");
+  } else if (capped.rowsCapped) {
+    const bigger = Math.min(AUDIT_MAX_LIMIT, limit * 5);
+    warnings.push(`Spans were capped: read ${capped.rows} ingest row(s), the --limit=${limit} maximum, so this trace ${capped.hasMoreKnown ? "has" : "may have"} more spans than the ${capped.spans} shown. ` + (capped.nextCursor ? "Page on with the next_cursor command, or " : "") + (limit < AUDIT_MAX_LIMIT ? `${capped.nextCursor ? "rerun" : "Rerun"} with --limit=${bigger} (max ${AUDIT_MAX_LIMIT}), or ${perTrace}.` : capped.nextCursor ? `${perTrace}.` : `${perTrace[0].toUpperCase()}${perTrace.slice(1)}.`));
+    if (limit < AUDIT_MAX_LIMIT && !params.cursor) {
+      nextCommands.push({ command: auditFlags([`--limit=${bigger}`]), purpose: "Read more ingest rows of this trace.", mutability: "read", requires_approval: false });
+    }
+  } else {
+    warnings.push(`Spans were capped: the server returned ${capped.spans} span(s) and marked the response truncated. ${perTrace[0].toUpperCase()}${perTrace.slice(1)}.`);
+  }
+  return {
+    data: { ...auditRecord, truncated: true, has_more: capped.rowsCapped, next_cursor: capped.nextCursor },
+    options: { warnings, ...nextCommands.length > 0 ? { nextCommands } : {} }
+  };
 }
 var TAIL_SEEN_MAX = 5000;
 var EMOTIONS_TAIL_PAGE_SIZE = 20;
@@ -9055,7 +9101,7 @@ async function readFullTranscript(conversationId, opts = {}) {
       out.push(`Message order may not be chronological: ${tied} of ${total} messages share a timestamp with another message ` + `(ordered by id, not time)${synthetic > 0 ? ` and ${synthetic} carry a synthetic pre-2022 timestamp` : ""}.`);
     }
     if (returned.some(isTruncated)) {
-      out.push(`Message text, tool-call inputs and tool results longer than ${CONTEXT_MESSAGE_TRUNCATE_CHARS} characters are truncated by the context endpoint. ` + "For full raw content (prompts, tool arguments/results) use `moda audit <conversation_id> --include-raw`.");
+      out.push(`Message text, tool-call inputs and tool results longer than ${CONTEXT_MESSAGE_TRUNCATE_CHARS} characters are truncated by the context endpoint. ` + "For full content (prompts, tool arguments/results) use `moda audit <conversation_id> --include-bodies` (or `--include-raw` for the verbatim events).");
     }
     return {
       data: stripLoneSurrogates({
@@ -9366,7 +9412,7 @@ async function runCommand(command, positional, flags, positionals = positional ?
     case "audit":
     case "trace": {
       if (!positional) {
-        throw new CliInputError("<conversation_id|trace_id> is required", "Usage: moda audit <conversation_id|trace_id> [--kind=auto|trace|conversation] [--include-raw] [--limit=N]");
+        throw new CliInputError("<conversation_id|trace_id> is required", "Usage: moda audit <conversation_id|trace_id> [--kind=auto|trace|conversation] [--include-raw] [--include-bodies] [--limit=N] [--cursor=<next_cursor>]");
       }
       const args = flagsToArgs(flags, "id", positional);
       const params = parseCommandFlags(AuditSchema, args, command, { positionalKey: "id" });
@@ -9375,36 +9421,26 @@ async function runCommand(command, positional, flags, positionals = positional ?
         query2.set("kind", params.kind);
       if (params.include_raw === true)
         query2.set("include_raw", "1");
+      if (params.include_bodies === true)
+        query2.set("include_bodies", "1");
       if (params.limit)
         query2.set("limit", params.limit.toString());
-      const queryString = query2.toString() ? `?${query2.toString()}` : "";
-      const data = await callDataAPI(`/audit/${encodeURIComponent(params.id)}${queryString}`);
-      const capped = auditCapStatus(data, params.limit ?? AUDIT_DEFAULT_LIMIT);
-      const auditRecord = asRecord(data);
-      if (!capped || !auditRecord) {
-        context.output.writeData(auditRecord ? { ...auditRecord, truncated: false, has_more: false } : data);
-        break;
-      }
-      const limit = params.limit ?? AUDIT_DEFAULT_LIMIT;
-      const perTrace = "audit one OTLP trace at a time with `moda audit <trace_id> --kind=trace` (ids in summary.trace_ids)";
-      const warnings = [];
-      const nextCommands = [];
-      if (capped.rowsCapped) {
-        const bigger = Math.min(AUDIT_MAX_LIMIT, limit * 5);
-        warnings.push(`Spans were capped: read ${capped.rows} ingest row(s), the --limit=${limit} maximum, so this trace ${capped.hasMoreKnown ? "has" : "may have"} more spans than the ${capped.spans} shown. ` + (limit < AUDIT_MAX_LIMIT ? `Rerun with --limit=${bigger} (max ${AUDIT_MAX_LIMIT}), or ${perTrace}.` : `${perTrace[0].toUpperCase()}${perTrace.slice(1)}.`));
-        if (limit < AUDIT_MAX_LIMIT) {
-          const rerun = [
-            `moda ${command} ${shellArg(params.id)}`,
-            params.kind && params.kind !== "auto" ? `--kind=${params.kind}` : "",
-            params.include_raw === true ? "--include-raw" : "",
-            `--limit=${bigger}`
-          ].filter(Boolean).join(" ");
-          nextCommands.push({ command: rerun, purpose: "Read more ingest rows of this trace.", mutability: "read", requires_approval: false });
-        }
-      } else {
-        warnings.push(`Spans were capped: the server returned ${capped.spans} span(s) and marked the response truncated. ${perTrace[0].toUpperCase()}${perTrace.slice(1)}.`);
+      if (params.cursor)
+        query2.set("cursor", params.cursor);
+      const agentMode = context.output.mode === "agent" || context.output.mode === "agent-stream";
+      let maxBytes = agentMode ? Math.max(AUDIT_MIN_PAGE_BYTES, agentMaxOutputBytes() - AUDIT_AGENT_RESERVE_BYTES) : undefined;
+      let page = auditOutput(command, params, await fetchAudit(params.id, query2, maxBytes));
+      while (maxBytes !== undefined && maxBytes > AUDIT_MIN_PAGE_BYTES && !agentEnvelopeFits(command, page.data, page.options)) {
+        maxBytes = Math.max(AUDIT_MIN_PAGE_BYTES, Math.floor(maxBytes / 2));
+        page = auditOutput(command, params, await fetchAudit(params.id, query2, maxBytes));
+      }
+      if (maxBytes !== undefined && !agentEnvelopeFits(command, page.data, page.options)) {
+        page.options.warnings = [
+          ...page.options.warnings ?? [],
+          "This page is a single ingest row larger than the agent output budget, so some of its spans are cut from this output. " + "Rerun this command with --json (no byte limit) or a higher MODA_AGENT_MAX_OUTPUT_BYTES to see the whole row before following next_cursor."
+        ];
       }
-      context.output.writeData({ ...auditRecord, truncated: true, has_more: capped.rowsCapped }, { warnings, ...nextCommands.length > 0 ? { nextCommands } : {} });
+      context.output.writeData(page.data, page.options);
       break;
     }
     case "frustrations": {
@@ -9483,11 +9519,7 @@ async function runCommand(command, positional, flags, positionals = positional ?
           trimmed = { ...trimmed, subtypes: ranked.slice(0, TOOL_FAILURE_AGENT_SUBTYPES), subtypes_total: subtypes.length };
           out.push(`Showing the ${TOOL_FAILURE_AGENT_SUBTYPES} most frequent of ${subtypes.length} failure subtypes; --json lists all of them.`);
         }
-        const { data: compacted, truncated } = compactToolFailureExamples(trimmed);
-        if (truncated > 0) {
-          out.push(`${truncated} long text field(s) in examples were cut to ${TOOL_FAILURE_TEXT_LIMIT} characters to keep this output small. For the full text, run each example's \`context_command\` (\`moda context <conversation_id> --msg-index=<msg_index>\`).`);
-        }
-        return compacted;
+        return trimmed;
       });
       break;
     }
@@ -10115,7 +10147,7 @@ var commandRegistry = createCommandRegistry([
     resetApiRequestCountBeforeRun: true,
     telemetry: "result",
     handler: async (context) => {
-      const { runInit } = await import("./index-p8vab4p4.js");
+      const { runInit } = await import("./index-z0jpjcsm.js");
       if (context.outputMode === "agent-stream") {
         context.output.writeEvent({
           event: "started",
@@ -10156,7 +10188,7 @@ var commandRegistry = createCommandRegistry([
     resetApiRequestCountBeforeRun: true,
     telemetry: "result-and-error",
     handler: async (context) => {
-      const { runProvision } = await import("./provision-e4cnxpdh.js");
+      const { runProvision } = await import("./provision-qkazpt0c.js");
       return runProvision(context, {
         resumeCommand: buildResumeCommand("provision", context.rawFlags)
       });
@@ -10633,7 +10665,9 @@ var commandRegistry = createCommandRegistry([
       "moda audit <conversation_id|trace_id>",
       "moda audit <trace_id> --kind=trace --json",
       "moda audit <conversation_id> --include-raw",
-      "moda audit <conversation_id> --limit=1000   # ingest rows to read (default 200, max 1000); warns when capped"
+      "moda audit <conversation_id> --include-bodies   # keep prompt/response bodies (omitted by default; long strings truncated)",
+      "moda audit <conversation_id> --limit=1000   # ingest rows to read (default 200, max 1000); warns when capped",
+      "moda audit <conversation_id> --cursor=<next_cursor>   # next page of a capped audit"
     ],
     ...dataApiDefaults
   }),
@@ -10651,7 +10685,7 @@ var commandRegistry = createCommandRegistry([
   }),
   legacyCommand({
     name: "tool-failure-detail",
-    description: "Get per-tool failure detail (each example embeds its anchor ±2 messages, long text cut to 300 chars)",
+    description: "Get per-tool failure detail (each example embeds its anchor ±2 messages)",
     examples: [
       "moda tool-failure-detail search",
       "moda tool-failure-detail search --compact   # no embedded conversation (same as --include-window=false)",
@@ -11061,6 +11095,15 @@ function printCliError(error) {
 ${issues.join(`
 `)}
 `);
+  } else if (error instanceof ApiError && error.readable) {
+    safeWriteStderr(`${error.message}
+`);
+    if (error.suggestions?.length) {
+      safeWriteStderr(`Try:
+${error.suggestions.map((s) => `  ${s}`).join(`
+`)}
+`);
+    }
   } else if (error instanceof ApiError) {
     safeWriteStderr(`API error (HTTP ${error.statusCode}): ${error.responseBody || error.message}
 `);

package/dist/{harness-vf1e1t2h.js → harness-7gr0xbtc.js}

@@ -42,8 +42,8 @@ import {
   writeHarnessAnalyzeResult,
   writeHarnessArtifacts,
   writeHumanProgress
-} from "./cli-tznhpmf2.js";
-import"./cli-233xm35b.js";
+} from "./cli-x0x0n35w.js";
+import"./cli-xrs02c0t.js";
 import"./cli-0v6na3yp.js";
 export {
   writeHumanProgress,

package/dist/{harness-github-actions-379faaqe.js → harness-github-actions-q9kewzb6.js}

@@ -1,7 +1,7 @@
 import {
   buildSourceSnapshot
-} from "./cli-tznhpmf2.js";
-import"./cli-233xm35b.js";
+} from "./cli-x0x0n35w.js";
+import"./cli-xrs02c0t.js";
 import"./cli-0v6na3yp.js";
 
 // src/harness-github-actions.ts

package/dist/{index-p8vab4p4.js → index-z0jpjcsm.js}

@@ -3,7 +3,7 @@ import {
   initPrompts,
   runPromptSync,
   runSkillSync
-} from "./cli-eaaegm7a.js";
+} from "./cli-jqfm8dp4.js";
 import {
   codingAgentDisplayName,
   describeCodingAgentEvent,
@@ -22,7 +22,7 @@ import {
   runHarnessCommand,
   startRemoteAnalyze,
   stripAnsi
-} from "./cli-tznhpmf2.js";
+} from "./cli-x0x0n35w.js";
 import {
   bundledSkillsDir,
   installIntegrationSkill,
@@ -34,14 +34,14 @@ import {
 } from "./cli-ssfq84k5.js";
 import {
   selectTenantAndCreateKey
-} from "./cli-ss2nere4.js";
+} from "./cli-h4k23fb0.js";
 import {
   ConnectCodeError,
   connectWithCode,
   loadAuthSession,
   loadToken,
   login
-} from "./cli-fpt3jkyh.js";
+} from "./cli-7kbzqact.js";
 import {
   CliAuthError,
   CliInputError,
@@ -49,7 +49,7 @@ import {
   createCommandContext,
   resolveIngestUrl,
   resolveModaBaseUrl
-} from "./cli-233xm35b.js";
+} from "./cli-xrs02c0t.js";
 import"./cli-0v6na3yp.js";
 
 // src/init/index.ts
@@ -604,6 +604,11 @@ function hasUnexpectedAuditDuplicates(value) {
     return new Set(calls).size !== calls.length;
   });
 }
+function promptHasSystemMessage(prompt) {
+  if (Array.isArray(prompt))
+    return prompt.some((message) => isRecord2(message) && message.role === "system");
+  return isRecord2(prompt) && prompt.reason === "row_payload_cap" && Array.isArray(prompt.roles) && prompt.roles.includes("system");
+}
 function completeAI7Projection(value) {
   if (!isAI7Audit(value))
     return true;
@@ -612,7 +617,7 @@ function completeAI7Projection(value) {
   const spans = value.spans.filter(isRecord2);
   for (const span of spans.filter((span2) => span2.type === "llm")) {
     if (isRecord2(span.attributes) && span.attributes["gen_ai.system_instructions"]) {
-      if (!Array.isArray(span.prompt) || !span.prompt.some((message) => isRecord2(message) && message.role === "system"))
+      if (!promptHasSystemMessage(span.prompt))
         return false;
     }
   }
@@ -632,7 +637,7 @@ function completeAI7Projection(value) {
   return expected.every((id) => typeof id === "string") && actual.every((id) => typeof id === "string") && new Set(actual).size === actual.length && expected.length === actual.length && expected.every((id) => actual.includes(id));
 }
 async function fetchAuditWithApiKey(id, apiKey, baseUrl) {
-  const url = `${baseUrl}/api/v1/data/audit/${encodeURIComponent(id)}?kind=conversation`;
+  const url = `${baseUrl}/api/v1/data/audit/${encodeURIComponent(id)}?kind=conversation&include_bodies=1`;
   const response = await fetch(url, {
     method: "GET",
     headers: { "x-api-key": apiKey },

package/dist/prompt-source-mcp.js

@@ -5,8 +5,8 @@ import {
   inspectCodeSelectors,
   inspectPythonSelectors,
   safePromptEvidencePath
-} from "./cli-tznhpmf2.js";
-import"./cli-233xm35b.js";
+} from "./cli-x0x0n35w.js";
+import"./cli-xrs02c0t.js";
 import {
   __commonJS,
   __toESM

package/dist/{provision-e4cnxpdh.js → provision-qkazpt0c.js}

@@ -1,15 +1,15 @@
 import {
   selectTenantAndCreateKey
-} from "./cli-ss2nere4.js";
+} from "./cli-h4k23fb0.js";
 import {
   loadAuthSession
-} from "./cli-fpt3jkyh.js";
+} from "./cli-7kbzqact.js";
 import {
   CliAuthError,
   resolveIngestUrl,
   resolveModaBaseUrl,
   stringOption
-} from "./cli-233xm35b.js";
+} from "./cli-xrs02c0t.js";
 import"./cli-0v6na3yp.js";
 
 // src/provision.ts

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@moda-ai/cli",
-  "version": "1.48.0",
+  "version": "1.50.0",
   "description": "CLI for Moda - AI agent analytics and observability",
   "type": "module",
   "bin": {

package/skills/integration/index.json

@@ -1,7 +1,7 @@
 {
   "schema_version": "moda.skill_index.v1",
-  "bundled_at": "2026-10-10T06:36:17.459Z",
-  "cli_version": "1.48.0",
+  "bundled_at": "2026-10-10T14:07:17.742Z",
+  "cli_version": "1.50.0",
   "skills": [
     {
       "id": "integration-cloudflare-think",

package/skills/moda-cli/SKILL.md

@@ -364,8 +364,8 @@ Fields an agent should read:
 
 `errors[].code` is one of `input_error`, `api_error`, `network_error`,
 `auth_required`, `input_required`, `cancelled`, `command_error`,
-`access_denied`, `unknown_error`, or a documented domain code such as
-`cloud_unavailable`. `access_denied` (exit 1) is an access problem — no
+`access_denied`, `output_error`, `unknown_error`, or a documented domain
+code such as `cloud_unavailable`. `access_denied` (exit 1) is an access problem — no
 organizations or an insufficient (viewer) role — distinct from `auth_required`
 (exit 4, missing/expired credentials).
 
@@ -408,14 +408,26 @@ Treat an answer as degraded if **any** of these hold:
 
 ### Large outputs and unknown flags
 
-- **Byte limit.** In `--agent` mode, envelopes over 200 KB are bounded
-  *without* breaking their shape: long text fields (usually embedded
-  transcripts) are shortened with `… [truncated N chars]`, then trailing list
-  items are dropped. `warnings[]` and `truncation.data` say exactly what was
-  cut (`dropped_items`, `complete: false`). `data` is still normal JSON, so
-  `data.frustrations[0].conversation_id` keeps working. For complete results,
-  page with a smaller `--limit` and `--offset`, or rerun with `--json` (no byte
-  limit).
+- **Byte limit.** In `--agent` mode, envelopes over 200 KB
+  (`MODA_AGENT_MAX_OUTPUT_BYTES`) are bounded *without* cutting any text or
+  breaking their shape: the complete output (the same JSON `--json` prints) is
+  written to a local file, and whole trailing list items are dropped from
+  `data`. `truncation.data.full_output_path` (also an `artifacts[]` entry of
+  kind `full_output`) is that file; read it in line ranges or with `jq`.
+  These files are pruned after 24 hours (`MODA_AGENT_OUTPUT_RETENTION_HOURS`).
+  `truncation.data` also says what was dropped (`dropped_items`,
+  `complete: false`). `data` is still normal JSON, so
+  `data.frustrations[0].conversation_id` keeps working. If even one item is
+  over budget, `data` is a stub `{ truncated, original_bytes, max_bytes,
+  full_output_path }`. If that file cannot be written (unwritable
+  `MODA_AGENT_OUTPUT_DIR`, full disk), nothing is dropped: the envelope keeps
+  the whole result over the limit with `status: "error"`, an `output_error`
+  (`cause: "full_output_save_failed"`), and exit `1`. If only the oversized
+  envelope text could not be saved (`cause: "envelope_text_save_failed"`),
+  that text stays whole, and the error message says whether `data` is whole
+  or bounded with its complete copy at `truncation.data.full_output_path`.
+  You can also page with a smaller `--limit` and
+  `--offset`, or rerun with `--json` (no byte limit).
 - **Unknown flags fail.** A flag the command doesn't take (a typo, or a
   filter it lacks) exits `1` with the valid flag list and a "did you mean",
   instead of being silently ignored. Stray arguments on commands that take
@@ -652,13 +664,23 @@ tool calls, prompt/response bodies, and `gen_ai.*` attributes — use `moda audi
 moda audit <conversation_id|trace_id>         # auto-detects OTLP trace_id vs trace ID (conversation_id)
 moda audit <trace_id> --kind=trace            # force OTLP trace_id lookup
 moda audit <conversation_id> --include-raw    # attach verbatim raw_event bodies
+moda audit <conversation_id> --include-bodies # keep prompt/response/attribute bodies
 moda audit <conversation_id> --limit=1000     # ingest rows to read (default 200, max 1000)
+moda audit <conversation_id> --cursor=<next_cursor>  # next page of a capped audit
 ```
 
 `truncated` / `has_more` at the top level say whether the audit is complete.
-A full `--limit` page of ingest rows (or a server span cap) sets both, and a
-warning plus `next_commands` give the rerun; for very large traces audit one
-OTLP trace at a time (`summary.trace_ids`, `--kind=trace`).
+A full `--limit` page of ingest rows, the server's ~3 MB response budget
+(`summary.truncated_reason: "byte_limit"`) or a span cap sets both, and a
+warning plus `next_commands` give the follow-up — usually the `--cursor=` page
+from `next_cursor`; for very large traces audit one OTLP trace at a time
+(`summary.trace_ids`, `--kind=trace`). Prompt/response bodies are omitted by
+default (`{"omitted": true, "original_bytes": N}`); `--include-bodies` keeps
+them, cutting strings over 8 KB to `{"truncated": true, "original_bytes": N,
+"preview": …}`. Ingest rows whose stored payload is too large carry
+`payload_capped: true`: their spans keep ids, hierarchy, timing and types but
+no bodies. A trace too large to audit at all fails fast with an
+`audit_too_large` error whose `suggested_commands` list per-trace audits.
 
 Returns the raw, NON-deduped span set from the Layer-0 store (`events_raw`), so
 nothing is collapsed. Use it to verify capture completeness: the `summary` gives
@@ -736,9 +758,9 @@ Pass `--include-window` to attach a `window` field per row (instead of
 (1–5, default `1`) is the half-width; `--window=1` ⇒ 3 messages total
 (anchor ± 1).
 
-Examples default to `--limit=5` (max 20). Strings inside `examples[]` longer
-than 300 characters are cut, and a warning says so; read the full text with
-the row's `context_command`. Agent output lists the 20 most frequent
+Examples default to `--limit=5` (max 20). Example text is returned whole; if
+the agent envelope goes over its byte limit, whole trailing examples are
+dropped and `truncation.data.full_output_path` holds every example. Agent output lists the 20 most frequent
 `subtypes` (`subtypes_total` has the full count); `--json` lists all.
 
 ### 7. Prompt management