aiden-runtime 4.1.0 → 4.1.1

package/README.md

@@ -1,3 +1,8 @@
+<img width="1672" height="941" alt="AIDEN BOOTUP LOGO" src="https://github.com/user-attachments/assets/c0809009-73e2-4d58-9292-12fbd0324952" />
+
+
+
+
 ```
 █████╗  ██╗██████╗ ███████╗███╗   ██╗
 ██╔══██╗██║██╔══██╗██╔════╝████╗  ██║
@@ -8,7 +13,7 @@
 
 Autonomous AI Engine
 
-19 providers · 68 skills · 42 tools · 9 channels · AGPL-3.0
+72 skills · 42 tools · 19 providers · 9 channels · AGPL-3.0
 
 Windows · Linux · WSL · macOS (API Mode)
 
@@ -16,12 +21,13 @@ Local-first · Self-healing routing · Browser & terminal control · Persistent
 ```
 
 <p align="center">
-  <a href="https://github.com/taracodlabs/aiden-releases/releases/latest"><img src="https://img.shields.io/github/v/release/taracodlabs/aiden-releases?color=f97316&label=version&style=for-the-badge" alt="Latest version" /></a>
-  <a href="https://github.com/taracodlabs/aiden-releases/releases"><img src="https://img.shields.io/github/downloads/taracodlabs/aiden-releases/total?color=f97316&label=downloads&style=for-the-badge" alt="Downloads" /></a>
+  <a href="https://github.com/taracodlabs/aiden/releases/latest"><img src="https://img.shields.io/github/v/release/taracodlabs/aiden?color=f97316&label=version&style=for-the-badge" alt="Latest version" /></a>
+  <a href="https://github.com/taracodlabs/aiden/releases"><img src="https://img.shields.io/github/downloads/taracodlabs/aiden/total?color=f97316&label=downloads&style=for-the-badge" alt="Downloads" /></a>
   <a href="https://discord.gg/gMZ3hUnQTm"><img src="https://img.shields.io/badge/chat-discord-7289da?logo=discord&logoColor=white&style=for-the-badge" alt="Discord" /></a>
   <a href="./LICENSE"><img src="https://img.shields.io/badge/license-AGPL--3.0-orange?style=for-the-badge" alt="License: AGPL-3.0" /></a>
   <a href="https://github.com/taracodlabs/aiden/stargazers"><img src="https://img.shields.io/github/stars/taracodlabs/aiden?style=for-the-badge&color=f9d71c" alt="Stars" /></a>
   <a href="https://www.npmjs.com/package/aiden-runtime"><img src="https://img.shields.io/npm/v/aiden-runtime?color=f97316&label=npm&style=for-the-badge" alt="npm" /></a>
+  <a href="https://amzn.to/4tpiXwM"><img src="https://img.shields.io/badge/book-Omega-ff9900?logo=amazon&logoColor=white&style=for-the-badge" alt="Book: Omega" /></a>
 </p>
 
 <p align="center">
@@ -103,14 +109,14 @@ Local-first · Self-healing routing · Browser & terminal control · Persistent
   <a href="https://aiden.taracod.com"><b>Website</b></a> &nbsp;·&nbsp;
   <a href="https://aiden.taracod.com/contact"><b>Contact</b></a> &nbsp;·&nbsp;
   <a href="https://discord.gg/gMZ3hUnQTm"><b>Discord</b></a> &nbsp;·&nbsp;
-  <a href="https://github.com/taracodlabs/aiden-releases/releases/latest"><b>Download</b></a> &nbsp;·&nbsp;
-  <a href="https://www.amazon.in/Omega-Shiva-Deore-ebook/dp/B0GX33VWZC/"><b>Book</b></a>
+  <a href="https://github.com/taracodlabs/aiden/releases/latest"><b>Download</b></a> &nbsp;·&nbsp;
+  <a href="https://amzn.to/4tpiXwM"><b>Book</b></a>
 </p>
 
 ---
 
-> **v4.1.0 — multi-channel autonomous engine · Telegram + MCP server + subagent fanout · voice CLI · hardened cron · skill mining · structured markdown rendering · cross-platform CI**
-> v4.1 turns Aiden into a multi-surface agent: a `ChannelAdapter`-shaped Telegram bot (text / voice / photo / PDF / groups / admin), an MCP server exposing 24 tools + the full skill catalog to Claude Desktop, parallel subagent fanout across `groq`/`together`, REPL voice mode (PTT + continuous), an auto-mining pipeline that proposes new skills from successful workflows, and a deep REPL polish layer (custom `@inquirer/core` prompt, autosuggest, sectioned boot card, sharp ASCII corners, theme detection). Linux/macOS/Windows × Node 20/22 CI matrix. See [changelog](#changelog) below.
+> **v4.1.0 — Multi-channel autonomous AI engine**
+> Telegram + MCP server + subagent fanout + voice CLI + skill mining. Hardened cron, structured markdown, cross-platform CI. See [changelog](#changelog) below.
 
 ---
 
@@ -326,7 +332,6 @@ We're shipping honest. Things that work, things that don't:
 
 **Not in v4.0:**
 
-- Subagent fanout / parallel agent swarm — single-loop only; deferred to v4.x
 - OCR — not bundled (vision-loop screen capture works, but no Tesseract)
 - Full agentskills.io ecosystem install — held pending license review
 - Docker sandbox backend — dropped in v4 rewrite
@@ -727,7 +732,6 @@ aiden                  # CLI
 
 - **npm package renamed** — `aiden-os` → `aiden-runtime`. Run `npm uninstall -g aiden-os && npm install -g aiden-runtime`.
 - **Slash commands consolidated** — v3's `/switch`, `/budget`, `/memory`, `/profile`, `/permissions`, `/sandbox`, `/retry`, `/failed`, `/publish` are gone. Use `/model`, `/usage`, `/identity`, `/yolo` for equivalent functionality. See `/help` for the v4 list.
-- **Subagent fanout removed** — v4 is single-loop only; subagent support deferred to v4.x.
 - **Docker sandbox dropped** — `AIDEN_SANDBOX_MODE` no longer applies. Tools run on the host. The `tirithScanner` secret/PII guard, `ssrfProtection`, and tiered approval engine remain as the safety layer.
 - **Skill registry install changed** — auto-fetch from external repos held pending license review. Skills install via `/skills install <local-path-or-url>` only at v4.0.
 - **Config compatible** — most environment variables (`OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `GROQ_API_KEY`, etc.) are recognised as-is. Copy your existing `.env` and Aiden picks them up.

package/dist/cli/v4/aidenCLI.js

@@ -282,12 +282,13 @@ async function main(argv, opts = {}) {
     program
         .command('doctor')
         .description('Run diagnostic checks')
-        .action(async () => {
+        .option('--providers', 'Also ping each configured / authed provider and report live status (deep check). Slower; useful before shipping or when a provider regression is suspected.')
+        .action(async (cmdOpts) => {
         if (opts.runDoctorHook) {
             await opts.runDoctorHook();
             return;
         }
-        await (0, doctor_1.runDoctorCli)();
+        await (0, doctor_1.runDoctorCli)({ liveness: cmdOpts.providers === true });
     });
     program
         .command('sessions <action> [arg]')

package/dist/cli/v4/doctor.js

@@ -18,6 +18,39 @@
  * will normally be sub-second.
  *
  */
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
 var __importDefault = (this && this.__importDefault) || function (mod) {
     return (mod && mod.__esModule) ? mod : { "default": mod };
 };
@@ -760,6 +793,33 @@ async function runDoctorCli(opts) {
         }
     }
     process.stdout.write(`\n${report.passed ? 'all checks passed' : 'some checks failed'} in ${report.totalMs} ms\n`);
-    process.exitCode = report.passed ? 0 : 1;
+    // Phase v4.1.1-oauth-fix Phase 5: discoverability hint for the deep
+    // mode. Only emitted in standard mode — if the user already passed
+    // `--providers`, the section right below this is the answer to the hint.
+    if (!opts?.liveness) {
+        process.stdout.write('  hint: Run `aiden doctor --providers` for live provider checks\n');
+    }
+    // Phase v4.1.1-oauth-fix Phase 4: opt-in provider liveness.
+    // Runs after the standard report so a failed config check is visible
+    // before the network probes start. Section header + summary line are
+    // rendered by doctorLiveness.renderProviderLivenessSection so the
+    // formatting stays alongside its consumer.
+    let livenessFailed = false;
+    if (opts?.liveness) {
+        process.stdout.write('\n  Running provider liveness checks...\n');
+        const { runProviderLiveness, renderProviderLivenessSection } = await Promise.resolve().then(() => __importStar(require('./doctorLiveness')));
+        const paths = opts.paths ?? (0, paths_1.resolveAidenPaths)();
+        const { results, summary } = await runProviderLiveness({
+            paths,
+            env: opts.env,
+            fetchImpl: opts.fetchImpl,
+            timeoutMs: opts.livenessTimeoutMs,
+        });
+        process.stdout.write(renderProviderLivenessSection(results, summary));
+        livenessFailed = summary.red > 0;
+    }
+    // Liveness reds count toward the overall exit code so CI / scripts
+    // can `aiden doctor --providers && deploy`.
+    process.exitCode = (report.passed && !livenessFailed) ? 0 : 1;
     return report;
 }

package/dist/cli/v4/doctorLiveness.js

@@ -0,0 +1,329 @@
+"use strict";
+/**
+ * Copyright (c) 2026 Shiva Deore (Taracod).
+ * Licensed under AGPL-3.0. See LICENSE for details.
+ *
+ * Aiden — local-first agent.
+ */
+/**
+ * cli/v4/doctorLiveness.ts — Phase v4.1.1-oauth-fix Phase 4.
+ *
+ * `aiden doctor --providers` deep-mode helper. Pings every configured /
+ * authed provider with a minimal request, reports green / red / skipped
+ * + per-provider latency + verbatim upstream error message.
+ *
+ * Why a separate file:
+ *   - Default doctor (`aiden doctor` with no flag) stays unchanged and
+ *     fast — config-shape checks only.
+ *   - `--providers` is opt-in. When the user types it we extend the
+ *     report with one liveness row per probe, then render a summary
+ *     line at the bottom.
+ *   - Tool-catalog validation is deliberately OUT of scope. Liveness
+ *     probes ship `tools: []` (see comment in checkProviderLiveness)
+ *     so one bad tool schema doesn't false-red every provider that
+ *     validates strictly. The eval-harness / registration-time schema
+ *     validator (v4.1.1 main) is the right home for that concern.
+ *
+ * Trust artifact:
+ *   - On failure we surface `err.message` VERBATIM (truncated to 200
+ *     chars). Phase 3 (`providers/v4/errors.ts`) already composes the
+ *     upstream response body into ProviderError.message — so a 400
+ *     prints the actual OpenAI reason, not a generic "provider failed."
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.enumerateConfiguredProviders = enumerateConfiguredProviders;
+exports.checkProviderLiveness = checkProviderLiveness;
+exports.runProviderLiveness = runProviderLiveness;
+exports.renderProviderLivenessSection = renderProviderLivenessSection;
+const registry_1 = require("../../providers/v4/registry");
+const runtimeResolver_1 = require("../../providers/v4/runtimeResolver");
+const credentialResolver_1 = require("../../providers/v4/credentialResolver");
+const tokenStore_1 = require("../../core/v4/auth/tokenStore");
+const DEFAULT_LIVENESS_TIMEOUT_MS = 8000;
+const PROBE_MAX_TOKENS = 4;
+const ERROR_TRUNCATE_CHARS = 200;
+const OLLAMA_PROBE_TIMEOUT_MS = 1500;
+const OLLAMA_HEALTH_URL = 'http://127.0.0.1:11434/api/tags';
+// ── Helpers ─────────────────────────────────────────────────────────────
+/**
+ * Truncate a long error string for display. The full error remains on
+ * the in-memory `LivenessResult.error` for programmatic consumers /
+ * test assertions.
+ */
+function truncate(s, max = ERROR_TRUNCATE_CHARS) {
+    if (s.length <= max)
+        return s;
+    return `${s.slice(0, max - 1)}…`;
+}
+/**
+ * Wrap a promise with a hard timeout. Resolves to the inner result on
+ * success, throws a clearly-labelled `Error` on timeout. Cleans up the
+ * timer either way.
+ */
+async function withTimeout(p, ms, label) {
+    let timer;
+    try {
+        return await Promise.race([
+            p,
+            new Promise((_, reject) => {
+                timer = setTimeout(() => reject(new Error(`${label}: timeout after ${ms}ms`)), ms);
+            }),
+        ]);
+    }
+    finally {
+        if (timer)
+            clearTimeout(timer);
+    }
+}
+/**
+ * Decide whether a provider counts as "configured" for liveness purposes.
+ * Three paths:
+ *   1. API key in env (`apiKeyEnvVar` set and the env var is non-empty).
+ *   2. OAuth (entry.oauth present and a non-expired token sits in the
+ *      tokenStore at <paths.root>/auth/<providerId>.json).
+ *   3. Local providers (ollama) — probed via a quick HTTP health check.
+ *
+ * Anything else is `configured: false` and gets a `skip_reason`.
+ */
+async function enumerateConfiguredProviders(opts) {
+    const env = opts.env ?? process.env;
+    const fetchImpl = opts.fetchImpl ?? fetch;
+    const out = [];
+    for (const entry of Object.values(registry_1.PROVIDER_REGISTRY)) {
+        // Every provider needs at least one model to probe against.
+        const model = entry.modelIds[0];
+        if (!model) {
+            out.push({
+                entry,
+                model: '',
+                configured: false,
+                reason: 'no models declared in registry',
+            });
+            continue;
+        }
+        // 1. API-key providers.
+        if (entry.apiKeyEnvVar) {
+            const value = env[entry.apiKeyEnvVar];
+            if (value && value.length > 0) {
+                out.push({ entry, model, configured: true });
+            }
+            else {
+                out.push({
+                    entry,
+                    model,
+                    configured: false,
+                    reason: `env ${entry.apiKeyEnvVar} not set`,
+                });
+            }
+            continue;
+        }
+        // 2. OAuth providers — check tokenStore.
+        if (entry.oauth) {
+            try {
+                const tokens = await (0, tokenStore_1.loadTokens)(opts.paths, entry.oauth.providerId);
+                if (tokens && tokens.accessToken) {
+                    if ((0, tokenStore_1.isExpired)(tokens, tokenStore_1.PREFLIGHT_REFRESH_WINDOW_MS)) {
+                        out.push({
+                            entry,
+                            model,
+                            configured: false,
+                            reason: 'OAuth token expired — run `/auth refresh` or `/auth login`',
+                        });
+                    }
+                    else {
+                        out.push({ entry, model, configured: true });
+                    }
+                }
+                else {
+                    out.push({
+                        entry,
+                        model,
+                        configured: false,
+                        reason: 'no OAuth token — run `/auth login`',
+                    });
+                }
+            }
+            catch (err) {
+                out.push({
+                    entry,
+                    model,
+                    configured: false,
+                    reason: `tokenStore read failed: ${err.message}`,
+                });
+            }
+            continue;
+        }
+        // 3. Local / no-credential providers (ollama). Configured = the
+        // local daemon answers a health probe.
+        const controller = new AbortController();
+        const timer = setTimeout(() => controller.abort(), OLLAMA_PROBE_TIMEOUT_MS);
+        try {
+            const res = await fetchImpl(OLLAMA_HEALTH_URL, { signal: controller.signal });
+            if (res.ok) {
+                out.push({ entry, model, configured: true });
+            }
+            else {
+                out.push({
+                    entry,
+                    model,
+                    configured: false,
+                    reason: `local daemon HTTP ${res.status}`,
+                });
+            }
+        }
+        catch (err) {
+            out.push({
+                entry,
+                model,
+                configured: false,
+                reason: `not running on ${OLLAMA_HEALTH_URL}`,
+            });
+        }
+        finally {
+            clearTimeout(timer);
+        }
+    }
+    return out;
+}
+/**
+ * Probe one provider/model pair via a minimal `adapter.call()`. Returns
+ * a structured result; never throws.
+ *
+ * On failure, `result.error` is `err.message` verbatim truncated to
+ * 200 chars. Phase v4.1.1-oauth-fix Phase 3 made `ProviderError.message`
+ * carry the upstream response body, so a 400 surfaces the actual
+ * reason (e.g. "Invalid schema for function 'subagent_fanout': …").
+ */
+async function checkProviderLiveness(provider, model, adapter, opts) {
+    const timeoutMs = opts?.timeoutMs ?? DEFAULT_LIVENESS_TIMEOUT_MS;
+    const start = Date.now();
+    // Liveness probes "is this provider reachable + authenticated?".
+    // Tool-catalog validation is a separate concern (eval harness,
+    // v4.1.1 main). Sending tools: [] ensures one bad tool schema
+    // doesn't false-red every provider that validates strictly.
+    const input = {
+        messages: [{ role: 'user', content: 'ping' }],
+        tools: [],
+        maxTokens: PROBE_MAX_TOKENS,
+    };
+    try {
+        await withTimeout(adapter.call(input), timeoutMs, `liveness ${provider}`);
+        return {
+            provider,
+            model,
+            status: 'green',
+            latency_ms: Date.now() - start,
+        };
+    }
+    catch (err) {
+        const raw = err instanceof Error ? err.message : String(err);
+        return {
+            provider,
+            model,
+            status: 'red',
+            latency_ms: Date.now() - start,
+            error: truncate(raw),
+        };
+    }
+}
+/**
+ * Run liveness probes against every provider returned by
+ * `enumerateConfiguredProviders`. Unconfigured providers come back as
+ * `status: 'skipped'` without any network traffic.
+ *
+ * Returns `{ results, summary }`. Summary tallies + a wall-clock total
+ * so the UI can print "X green · Y red · Z skipped · NNNms total".
+ */
+async function runProviderLiveness(opts) {
+    const start = Date.now();
+    const parallel = opts.parallel ?? true;
+    const resolver = opts.resolverImpl ?? new runtimeResolver_1.RuntimeResolver(new credentialResolver_1.CredentialResolver(opts.paths.authJson));
+    const configured = await enumerateConfiguredProviders({
+        paths: opts.paths,
+        env: opts.env,
+        fetchImpl: opts.fetchImpl,
+    });
+    const probe = async (c) => {
+        if (!c.configured) {
+            return {
+                provider: c.entry.id,
+                status: 'skipped',
+                latency_ms: 0,
+                skip_reason: c.reason ?? 'not configured',
+            };
+        }
+        try {
+            const adapter = await resolver.resolve({
+                providerId: c.entry.id,
+                modelId: c.model,
+                paths: opts.paths,
+            });
+            return await checkProviderLiveness(c.entry.id, c.model, adapter, { timeoutMs: opts.timeoutMs });
+        }
+        catch (err) {
+            // Resolve failure (missing credential, unknown model, etc.).
+            // Same surface treatment as a probe failure.
+            const raw = err instanceof Error ? err.message : String(err);
+            return {
+                provider: c.entry.id,
+                model: c.model,
+                status: 'red',
+                latency_ms: 0,
+                error: truncate(raw),
+            };
+        }
+    };
+    const results = parallel
+        ? await Promise.all(configured.map(probe))
+        : await runSequential(configured.map((c) => () => probe(c)));
+    const summary = {
+        green: results.filter((r) => r.status === 'green').length,
+        red: results.filter((r) => r.status === 'red').length,
+        skipped: results.filter((r) => r.status === 'skipped').length,
+        total_ms: Date.now() - start,
+    };
+    return { results, summary };
+}
+async function runSequential(thunks) {
+    const out = [];
+    for (const t of thunks)
+        out.push(await t());
+    return out;
+}
+/**
+ * Render the liveness section as plain-text rows. The doctor command
+ * prints this BELOW the standard health box so the default `aiden doctor`
+ * output stays byte-identical.
+ *
+ * Visual style matches the existing doctor rows (✓ / ✗ / -) but lays
+ * out as a tabular block rather than a box — the rows can be long
+ * (upstream error bodies) and forcing them into the 100-col box would
+ * truncate the diagnostic that's the whole point of the feature.
+ */
+function renderProviderLivenessSection(results, summary) {
+    const nameWidth = Math.max(8, ...results.map((r) => r.provider.length));
+    const lines = [];
+    lines.push('');
+    lines.push('  Provider liveness (deep check)');
+    lines.push(`  ${'─'.repeat(60)}`);
+    for (const r of results) {
+        const icon = r.status === 'green' ? '✓'
+            : r.status === 'red' ? '✗'
+                : '-';
+        const name = r.provider.padEnd(nameWidth);
+        const status = r.status === 'green' ? 'green'.padEnd(8)
+            : r.status === 'red' ? 'red  '.padEnd(8)
+                : 'skip '.padEnd(8);
+        const latency = r.latency_ms > 0
+            ? `${r.latency_ms}ms`.padEnd(8)
+            : ''.padEnd(8);
+        const tail = r.status === 'green' ? (r.model ?? '')
+            : r.status === 'red' ? (r.error ?? 'unknown error')
+                : (r.skip_reason ?? 'not configured');
+        lines.push(`  ${icon} ${name}  ${status}${latency}${tail}`);
+    }
+    lines.push(`  ${'─'.repeat(60)}`);
+    lines.push(`  ${summary.green} green · ${summary.red} red · ${summary.skipped} skipped · ${summary.total_ms}ms total`);
+    lines.push('');
+    return lines.join('\n');
+}

package/dist/core/v4/skillMining/extractorPrompt.js

@@ -16,10 +16,10 @@
  *      metadata.aiden.* must round-trip unchanged. We re-parse the
  *      refined output and discard it if any required field drifts.
  *
- *   2. Never write attribution tokens (hermes / nous / "portions
- *      adapted from" / "original copyright"). The permanent
- *      attribution sweep validates this; if a refined output
- *      contains any forbidden token, we fall back to the skeleton.
+ *   2. Never write attribution tokens or "portions adapted from..." /
+ *      "original copyright" strings. The permanent attribution
+ *      sweep validates this; if a refined output contains any
+ *      forbidden token, we fall back to the skeleton.
  *
  * If the auxiliary client is unavailable, the call times out, or
  * the refined output fails validation, the function returns the
@@ -29,23 +29,30 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.refine = refine;
 const skillSpec_1 = require("../skillSpec");
-const FORBIDDEN_TOKENS_RE = /\b(hermes|nous|portions adapted from|original copyright)\b/i;
-const REFINER_SYSTEM_PROMPT = `
-You polish auto-generated skill markdown for a local-first AI agent.
-
-Your job is to improve the WORDING ONLY of an already-valid SKILL.md
-file. The frontmatter (everything between the leading "---" markers)
-must round-trip BYTE-FOR-BYTE unchanged. The "# <name>" heading must
-stay first in the body. The numbered "## Steps" list must remain
-numbered and in the same order; you may rephrase step descriptions
-but must NOT add or remove steps.
-
-Hard rules:
-- Output the COMPLETE SKILL.md, not a diff.
-- Do not add boilerplate, citations, or attribution to any other
-  agent or codebase. The skill is 100% the user's own.
-- Do not introduce mock/fake values into commands.
-- Keep total length under 6000 characters.
+const BANNED_TOKENS = [
+    'portions adapted from',
+    'original copyright',
+    'derived from',
+    'based on the',
+    'adapted from',
+];
+const FORBIDDEN_TOKENS_RE = new RegExp(`\\b(${BANNED_TOKENS.join('|')})\\b`, 'i');
+const REFINER_SYSTEM_PROMPT = `
+You polish auto-generated skill markdown for a local-first AI agent.
+
+Your job is to improve the WORDING ONLY of an already-valid SKILL.md
+file. The frontmatter (everything between the leading "---" markers)
+must round-trip BYTE-FOR-BYTE unchanged. The "# <name>" heading must
+stay first in the body. The numbered "## Steps" list must remain
+numbered and in the same order; you may rephrase step descriptions
+but must NOT add or remove steps.
+
+Hard rules:
+- Output the COMPLETE SKILL.md, not a diff.
+- Do not add boilerplate, citations, or attribution to any other
+  agent or codebase. The skill is 100% the user's own.
+- Do not introduce mock/fake values into commands.
+- Keep total length under 6000 characters.
 `.trim();
 /**
  * Refine `skeleton` via the auxiliary client. Always returns a

package/dist/core/v4/skillMining/proposalBuilder.js

@@ -18,8 +18,9 @@
  *   - emits required fields `name`, `description`, `version`
  *   - emits `metadata.aiden` with the mining provenance fields
  *   - body is numbered tool-call steps in markdown
- *   - never writes hermes/nous/copyright attribution strings — the
- *     permanent attribution sweep validates this
+ *   - never writes attribution strings — the banned-token regex
+ *     strips them at extraction time and the permanent attribution
+ *     sweep validates the result
  */
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.deriveName = deriveName;

package/dist/core/version.js

@@ -2,4 +2,4 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.VERSION = void 0;
 // AUTO-GENERATED by scripts/inject-version.js — do not edit by hand
-exports.VERSION = '4.1.0';
+exports.VERSION = '4.1.1';

package/dist/moat/dangerousPatterns.js

@@ -52,7 +52,7 @@ exports.DANGEROUS_PATTERNS = [
     { name: 'kill_all', regex: /\bkill\s+-9\s+-1\b/, tier: 'dangerous', description: 'kill all processes' },
     { name: 'pkill_force', regex: /\bpkill\s+-9\b/, tier: 'caution', description: 'force kill processes' },
     { name: 'systemctl_disable', regex: /\bsystemctl\s+(-[^\s]+\s+)*(stop|restart|disable|mask)\b/i, tier: 'caution', description: 'stop/restart system service' },
-    { name: 'pkill_aiden', regex: /\b(pkill|killall)\b.*\b(aiden|gateway|hermes)\b/i, tier: 'dangerous', description: 'kill aiden/gateway process (self-termination)' },
+    { name: 'pkill_aiden', regex: /\b(pkill|killall)\b.*\b(aiden|gateway)\b/i, tier: 'dangerous', description: 'kill aiden/gateway process (self-termination)' },
     // ── Sensitive write targets ───────────────────────────────────
     { name: 'write_etc', regex: />\s*\/etc\//, tier: 'dangerous', description: 'overwrite system config' },
     { name: 'tee_etc', regex: /\btee\b.*\/etc\//, tier: 'dangerous', description: 'overwrite system file via tee' },

package/dist/providers/v4/codexResponsesAdapter.js

@@ -129,14 +129,19 @@ class CodexResponsesAdapter {
         const body = {
             model: this.model,
             input: items,
-            tool_choice: 'auto',
-            parallel_tool_calls: true,
             store: false,
         };
         if (instructions)
             body.instructions = instructions;
+        // Phase v4.1.1-oauth-fix Phase 5: `tool_choice` and
+        // `parallel_tool_calls` are only meaningful when tools are present.
+        // OpenAI Codex returns HTTP 400 (empty body) for `tool_choice: 'auto'`
+        // without a `tools` field — surfaced by `aiden doctor --providers`'s
+        // no-tools liveness probe.
         if (input.tools && input.tools.length > 0) {
             body.tools = input.tools.map(toWireTool);
+            body.tool_choice = 'auto';
+            body.parallel_tool_calls = true;
         }
         if (typeof input.temperature === 'number') {
             body.temperature = input.temperature;

package/dist/providers/v4/errors.js

@@ -16,9 +16,59 @@
  */
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.ProviderRateLimitError = exports.ProviderTimeoutError = exports.ProviderError = void 0;
+exports.formatRawForMessage = formatRawForMessage;
+/**
+ * Format a raw response body for inclusion in the user-facing error
+ * message. Recognises the OpenAI / Anthropic JSON envelope shape
+ * (`{ error: { message: "..." } }`) and falls back to the raw string
+ * for plain-text bodies. Returns null when nothing useful is available
+ * so callers can omit the ": <detail>" tail entirely.
+ *
+ * Truncates to 300 chars to keep multi-line responses from blowing
+ * up the user's terminal — full body remains on `error.raw` for
+ * programmatic consumers / `aiden doctor --providers` deep mode.
+ */
+function formatRawForMessage(raw) {
+    if (raw === undefined || raw === null)
+        return null;
+    // OpenAI / Anthropic JSON envelope: { error: { message: "..." } }
+    if (typeof raw === 'object') {
+        const err = raw.error;
+        if (err && typeof err === 'object') {
+            const msg = err.message;
+            if (typeof msg === 'string' && msg.length > 0) {
+                return msg.length > 300 ? `${msg.slice(0, 300)}…` : msg;
+            }
+        }
+        // Some providers put the message at the top level.
+        const topMsg = raw.message;
+        if (typeof topMsg === 'string' && topMsg.length > 0) {
+            return topMsg.length > 300 ? `${topMsg.slice(0, 300)}…` : topMsg;
+        }
+        return null;
+    }
+    // Plain string body.
+    if (typeof raw === 'string') {
+        const trimmed = raw.trim();
+        if (trimmed.length === 0)
+            return null;
+        return trimmed.length > 300 ? `${trimmed.slice(0, 300)}…` : trimmed;
+    }
+    return null;
+}
+/**
+ * Compose the final `Error.message` from the short summary and (when
+ * available) the parsed/truncated raw response body. The body remains
+ * stashed on `ProviderError.raw` either way — this only enriches what
+ * users see when the error is rendered.
+ */
+function composeMessage(message, raw) {
+    const tail = formatRawForMessage(raw);
+    return tail ? `${message}: ${tail}` : message;
+}
 class ProviderError extends Error {
     constructor(message, providerName, statusCode, raw, retryable = false) {
-        super(message);
+        super(composeMessage(message, raw));
         this.providerName = providerName;
         this.statusCode = statusCode;
         this.raw = raw;

package/dist/providers/v4/ollamaPromptToolsAdapter.js

@@ -71,7 +71,12 @@ class OllamaPromptToolsAdapter {
                 const status = response.status;
                 const rawText = await this.safeReadText(response);
                 const retryable = status >= 500 || status === 429;
-                const err = new errors_1.ProviderError(`Provider ${this.providerName} returned ${status}: ${rawText.slice(0, 500)}`, this.providerName, status, rawText, retryable);
+                // Phase v4.1.1-oauth-fix Phase 5: short message only. The raw
+                // body flows via the .raw arg and composeMessage in errors.ts
+                // appends a truncated summary into the final .message — this
+                // file used to inline it, producing duplicated output in
+                // `aiden doctor --providers` and anywhere else err.message is logged.
+                const err = new errors_1.ProviderError(`Provider ${this.providerName} returned ${status}`, this.providerName, status, rawText, retryable);
                 if (!retryable || attempt >= totalAttempts)
                     throw err;
                 lastError = err;
@@ -149,7 +154,9 @@ class OllamaPromptToolsAdapter {
             clearTimeout(timer);
             const status = response.status;
             const rawText = await this.safeReadText(response);
-            throw new errors_1.ProviderError(`Provider ${this.providerName} returned ${status}: ${rawText.slice(0, 500)}`, this.providerName, status, rawText, status >= 500);
+            // Phase v4.1.1-oauth-fix Phase 5: composeMessage handles body
+            // rendering centrally; inlining it here would duplicate.
+            throw new errors_1.ProviderError(`Provider ${this.providerName} returned ${status}`, this.providerName, status, rawText, status >= 500);
         }
         if (!response.body) {
             clearTimeout(timer);

package/dist/tools/v4/subagent/subagentFanout.js

@@ -73,6 +73,30 @@ function makeSubagentFanoutTool(factory) {
                     tasks: {
                         type: 'array',
                         description: 'Per-child task list (partition mode only). Length must equal n.',
+                        // Schema mirrors PartitionTask interface in
+                        // core/v4/subagent/fanout.ts:70-75. If you change one, change
+                        // the other. OpenAI Codex backend strictly validates schemas
+                        // and rejects `type: "array"` declarations missing `items`,
+                        // so the inner shape must be explicit here.
+                        items: {
+                            type: 'object',
+                            description: 'One unit of work for a partition-mode child.',
+                            properties: {
+                                goal: {
+                                    type: 'string',
+                                    description: 'The task this child should accomplish.',
+                                },
+                                context: {
+                                    type: 'string',
+                                    description: 'Optional shared context for the child.',
+                                },
+                                role: {
+                                    type: 'string',
+                                    description: 'Optional role tag, diagnostic only.',
+                                },
+                            },
+                            required: ['goal'],
+                        },
                     },
                     merge: {
                         type: 'string',

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "aiden-runtime",
-  "version": "4.1.0",
+  "version": "4.1.1",
   "publishConfig": {
     "access": "public"
   },