discoclaw 0.4.0 → 0.5.1

package/.context/discord.md

@@ -86,12 +86,15 @@ Each action category has its own flag (only active when the master switch is `1`
 | `DISCOCLAW_DISCORD_ACTIONS_MEMORY` | `1` | memoryRemember, memoryForget, memoryShow |
 | `DISCOCLAW_DISCORD_ACTIONS_DEFER` | `1` | defer |
 | `DISCOCLAW_DISCORD_ACTIONS_IMAGEGEN` | `0` | generateImage |
+| `DISCOCLAW_DISCORD_ACTIONS_VOICE` | `0` | voiceStatus, voiceJoin, voiceLeave, voiceSetVoice |
+| `DISCOCLAW_DISCORD_ACTIONS_SPAWN` | `1` | spawnAgent |
 | _(config — always on)_ | — | modelSet, modelShow |
 
 Notes:
 - `reactionPrompt` is gated by the MESSAGING flag — it is registered via `REACTION_PROMPT_ACTION_TYPES` only when `flags.messaging` is true (`src/discord/actions.ts:113`).
 - Config actions (`modelSet`, `modelShow`) have no separate env flag. They are always enabled when the master switch is on, hardcoded in `src/index.ts`.
 - `generateImage` supports two providers: **OpenAI** (models: `dall-e-3`, `gpt-image-1`) and **Gemini** (models: `imagen-4.0-generate-001`, `imagen-4.0-fast-generate-001`, `imagen-4.0-ultra-generate-001`). Provider is auto-detected from the model prefix (`dall-e-*`/`gpt-image-*` → openai, `imagen-*` → gemini) or set explicitly via the `provider` field. OpenAI provider uses `OPENAI_API_KEY` (required) and optional `OPENAI_BASE_URL`. Gemini provider uses `IMAGEGEN_GEMINI_API_KEY`. At least one key must be set when `DISCOCLAW_DISCORD_ACTIONS_IMAGEGEN=1`. Default model is auto-detected: if only `IMAGEGEN_GEMINI_API_KEY` is set, defaults to `imagen-4.0-generate-001`; otherwise defaults to `dall-e-3`. Override with `IMAGEGEN_DEFAULT_MODEL`.
+- `spawnAgent` is enabled by default (`DISCOCLAW_DISCORD_ACTIONS_SPAWN=1`; set to 0 to disable). Spawned agents run fire-and-forget: each agent runs its prompt via the configured runtime and posts its output directly to the target channel. Multiple `spawnAgent` actions in a single response run in parallel (bounded by `DISCOCLAW_DISCORD_ACTIONS_SPAWN_MAX_CONCURRENT`, default 8). Spawn is disabled for bot-originated messages and excluded from cron flows to prevent recursive agent chains. Spawned agents run at recursion depth 1 and cannot themselves spawn further agents.
 
 Auto-follow-up: When query actions (channelList, channelInfo, threadListArchived, forumTagList, readMessages, fetchMessage, listPins, memberInfo, roleInfo, searchMessages, eventList, taskList, taskShow, cronList, cronShow, planList, planShow, memoryShow, modelShow, forgeStatus) succeed, DiscoClaw automatically re-invokes Claude with the results. This allows Claude to reason about query results without requiring the user to send a follow-up message. Controlled by `DISCOCLAW_ACTION_FOLLOWUP_DEPTH` (default `3`, `0` disables). Mutation-only responses do not trigger follow-ups. Trivially short follow-up responses (<50 chars with no actions) are suppressed.
 

package/.context/memory.md

@@ -187,13 +187,14 @@ no separator). The three memory builders run in `Promise.all` so they add no lat
 | Rolling summary | 2000 chars | on | The `fast`-tier model is prompted with `"Keep the summary under {maxChars} characters"`. Replaces itself each update rather than growing. |
 | Message history | 3000 chars | on | Fetches up to 10 messages, walks backward from newest. Bot messages truncated to fit; user messages that don't fit cause a hard stop. |
 | Short-term memory | 1000 chars | **on** | Filters by max age (default 6h), sorts newest-first, accumulates lines until budget hit. |
+| Open tasks | 600 chars | **on** | Queries TaskStore for non-closed tasks at invocation time, formats as a compact list. |
 | Auto-extraction | n/a | **off** | Write-side only — extracts facts for future prompts, adds nothing to the current turn. |
 | Workspace files | no budget | on (DMs only) | Loaded as file paths, not inlined. The runtime reads them on demand. |
 
 ### Default prompt overhead
 
-With the three enabled layers at default settings, worst-case memory overhead is
-**~7000 chars (~1750 tokens)**. With all layers enabled, ~8000 chars (~2000 tokens).
+With the four enabled layers at default settings, worst-case memory overhead is
+**~8600 chars (~2150 tokens)**.
 This is modest against typical `capable`-tier context windows.
 
 In practice most prompts use far less — a user with 5 durable items and a short summary
@@ -216,6 +217,7 @@ Memory sections are injected into every prompt in this order:
 Context files (PA + MEMORY.md + daily logs + channel context)
   → Durable memory section (up to 2000 chars)
   → Short-term memory section (up to 1000 chars)
+  → Open tasks section (up to 600 chars)
   → Rolling summary section (up to 2000 chars)
   → Message history (up to 3000 chars)
   → Discord actions

package/.context/pa.md

@@ -26,6 +26,7 @@ Templates live in `templates/workspace/` and are scaffolded on first run (copy-i
 - **Never go silent.** Acknowledge before tool calls.
 - Narrate failures and pivots.
 - Summarize outcomes; don't assume the user saw tool output.
+- **Close the loop.** When query actions return data, answer the user's original question in the same response. Don't chain query after query without delivering a conclusion. If you need more data, say what you're checking and why — but every auto-follow-up response should either deliver the answer or visibly advance toward one.
 - **Never edit `tasks.jsonl`, cron store files, or other data files directly.** Always use the corresponding discord action (`taskUpdate`, `taskCreate`, `cronUpdate`, etc.). Direct file edits bypass Discord thread sync and leave the UI stale.
 
 ## Discord Formatting

package/.context/tasks.md

@@ -5,6 +5,13 @@ Tasks are backed by an in-process `TaskStore` and synced to Discord forum thread
 Ground-zero post-hard-cut refactor tracker: `docs/tasks-ground-zero-post-hard-cut-plan.md`
 Ground-zero task architecture refactor status: COMPLETE (FROZEN), verified on 2026-02-21.
 
+## Prompt Injection
+
+Open task statuses are injected into prompts at invocation time, sourced directly
+from the TaskStore. This ensures every new session starts with accurate task state
+regardless of rolling summary freshness. See `.context/memory.md` for assembly order
+and token budget.
+
 ## Data Model
 
 Canonical type: `TaskData` in `src/tasks/types.ts`.

package/.context/voice.md

@@ -83,5 +83,6 @@ When `voiceEnabled=true`, the post-connect block in `src/index.ts` initializes t
 | `DEEPGRAM_API_KEY` | — | Required for deepgram STT and TTS |
 | `DEEPGRAM_STT_MODEL` | `nova-3-conversationalai` | Deepgram STT model name |
 | `DEEPGRAM_TTS_VOICE` | `aura-2-asteria-en` | Deepgram TTS voice name |
+| `DEEPGRAM_TTS_SPEED` | `1.3` | Deepgram TTS playback speed (range 0.5–1.5) |
 | `CARTESIA_API_KEY` | — | Required for cartesia TTS |
 | *(built-in)* | — | Telegraphic style instruction hardcoded into every voice AI invocation — front-loads the answer, strips preambles/markdown/filler, keeps responses short for TTS latency. Not an env var; not overridable by `DISCOCLAW_VOICE_SYSTEM_PROMPT`. |

package/.env.example

@@ -32,7 +32,7 @@ DISCORD_ALLOW_USER_IDS=
 # connect and persisted to system-scaffold.json. Only set this to override the auto-created channel.
 #DISCOCLAW_CRON_FORUM=
 
-# Default model for cron job execution: fast | capable (or concrete model names).
+# Default model for cron job execution: fast | capable | deep (or concrete model names).
 # Defaults to capable (resolves to Sonnet on Claude Code) — avoids using Opus on routine cron work.
 # Override at runtime via `!models set cron-exec <model>`.
 #DISCOCLAW_CRON_EXEC_MODEL=capable
@@ -62,7 +62,7 @@ DISCORD_GUILD_ID=
 # Disable Codex session persistence/resume (workaround for session DB issues):
 #CODEX_DISABLE_SESSIONS=1
 
-# Model tier: fast | capable (provider-agnostic).
+# Model tier: fast | capable | deep (provider-agnostic).
 # Concrete model names (e.g. opus, sonnet, gpt-4o) are still accepted as passthrough.
 #RUNTIME_MODEL=capable
 

package/.env.example.full

@@ -49,36 +49,40 @@ DISCORD_ALLOW_USER_IDS=
 #PRIMARY_RUNTIME=claude
 
 # --- Primary models ---
-# Model tier: fast | capable (provider-agnostic).
+# Model tier: fast | capable | deep (provider-agnostic).
 # Concrete model names (e.g. opus, sonnet, gpt-4o) are still accepted as passthrough.
 #RUNTIME_MODEL=capable
 
 # --- Fast-tier default ---
 # Sets the default model for all "small" tasks (summary, cron, cron auto-tag, task auto-tag).
-# Individual overrides (DISCOCLAW_SUMMARY_MODEL, etc.) still win when set.
+# Valid tiers: fast | capable | deep. Individual overrides (DISCOCLAW_SUMMARY_MODEL, etc.) still win when set.
 #DISCOCLAW_FAST_MODEL=fast
 
 # --- Tier model overrides ---
 # Override the concrete model resolved for any runtime × tier combination.
-# Format: DISCOCLAW_TIER_<RUNTIME>_<FAST|CAPABLE>=<model>
+# Format: DISCOCLAW_TIER_<RUNTIME>_<FAST|CAPABLE|DEEP>=<model>
 # Unset = use the built-in default shown in the comments.
 # Concrete model names (e.g. sonnet, gpt-4o-mini) are passed through unchanged.
 #
 # Claude Code adapter (default: fast=haiku, capable=sonnet):
 #DISCOCLAW_TIER_CLAUDE_CODE_FAST=haiku
 #DISCOCLAW_TIER_CLAUDE_CODE_CAPABLE=sonnet
+#DISCOCLAW_TIER_CLAUDE_CODE_DEEP=claude-opus-4-6
 #
 # Gemini CLI adapter (default: fast=gemini-2.5-flash, capable=gemini-2.5-pro):
 #DISCOCLAW_TIER_GEMINI_FAST=gemini-2.5-flash
 #DISCOCLAW_TIER_GEMINI_CAPABLE=gemini-2.5-pro
+#DISCOCLAW_TIER_GEMINI_DEEP=gemini-2.5-pro
 #
 # OpenAI-compatible adapter (default: adapter-default for both tiers):
 #DISCOCLAW_TIER_OPENAI_FAST=gpt-4o-mini
 #DISCOCLAW_TIER_OPENAI_CAPABLE=gpt-4o
+#DISCOCLAW_TIER_OPENAI_DEEP=
 #
 # Codex CLI adapter (default: adapter-default for both tiers):
 #DISCOCLAW_TIER_CODEX_FAST=
 #DISCOCLAW_TIER_CODEX_CAPABLE=
+#DISCOCLAW_TIER_CODEX_DEEP=
 #
 # Tool-tier map — override which tool tier a model resolves to.
 # Comma-separated model=tier pairs. Tiers: basic, standard, full.
@@ -111,9 +115,9 @@ DISCORD_ALLOW_USER_IDS=
 # Cron — forum-based scheduled tasks (enabled by default)
 # ----------------------------------------------------------
 # Forum channel ID is auto-created on first connect (see AUTO-DETECTED above).
-# Model tier for cron execution: fast | capable (concrete names accepted as passthrough).
+# Model tier for cron execution: fast | capable | deep (concrete names accepted as passthrough).
 #DISCOCLAW_CRON_MODEL=fast
-# Default model tier for cron job execution (fast | capable, concrete names accepted as passthrough).
+# Default model tier for cron job execution (fast | capable | deep, concrete names accepted as passthrough).
 # Defaults to capable (resolves to Sonnet on Claude Code) — avoids using Opus on routine cron work.
 # Per-job overrides and AI-classified model still win when set. Override at runtime via
 # `!models set cron-exec <model>`.
@@ -197,6 +201,14 @@ DISCOCLAW_DISCORD_ACTIONS_DEFER=1
 # Maximum number of pending deferred invocations allowed at once (default: 5).
 #DISCOCLAW_DISCORD_ACTIONS_DEFER_MAX_CONCURRENT=5
 
+# Allow the AI to spawn parallel sub-agents in target channels.
+# Each spawned agent is a fire-and-forget invocation that posts output to the specified channel.
+# Spawned agents run at recursion depth 1 and cannot themselves spawn further agents.
+# Default: on. Set to 0 to disable.
+#DISCOCLAW_DISCORD_ACTIONS_SPAWN=1
+# Maximum number of spawn actions that may run concurrently (default: 8).
+#DISCOCLAW_DISCORD_ACTIONS_SPAWN_MAX_CONCURRENT=8
+
 # ============================================================
 # OPTIONAL — uncomment sections you need
 # ============================================================
@@ -323,7 +335,7 @@ DISCOCLAW_DISCORD_ACTIONS_DEFER=1
 #USE_GROUP_DIR_CWD=0
 # Tools available to the runtime (Glob, Grep, Write now included by default).
 #RUNTIME_TOOLS=Bash,Read,Write,Edit,Glob,Grep,WebSearch,WebFetch
-# Fallback model tier when primary is overloaded (concrete names accepted as passthrough).
+# Fallback model tier when primary is overloaded: fast | capable | deep (concrete names accepted as passthrough).
 #RUNTIME_FALLBACK_MODEL=fast
 # Max USD spend per Claude CLI process. One-shot: per invocation.
 # Multi-turn: per session lifetime (accumulates across turns). Recommend $5-10.
@@ -521,7 +533,7 @@ DISCOCLAW_DISCORD_ACTIONS_IMAGEGEN=0
 # in system-scaffold.json). Only set this to override the auto-discovered channel.
 # Leave unset to disable transcript mirroring.
 #DISCOCLAW_VOICE_LOG_CHANNEL=  # e.g. "voice-log" if using the default scaffold
-# Model for voice AI responses: tier (fast | capable) or concrete name (sonnet, opus, haiku).
+# Model for voice AI responses: tier (fast | capable | deep) or concrete name (sonnet, opus, haiku).
 # Independent of RUNTIME_MODEL — allows tuning voice latency vs quality separately from chat.
 # Switchable at runtime via `modelSet voice <model>`.
 # Default: follows DISCOCLAW_FAST_MODEL (override here for voice-specific tuning).
@@ -537,5 +549,8 @@ DISCOCLAW_DISCORD_ACTIONS_IMAGEGEN=0
 # Deepgram TTS voice for speech synthesis (default: aura-2-asteria-en).
 # See https://developers.deepgram.com/docs/tts-models for available voices.
 #DEEPGRAM_TTS_VOICE=aura-2-asteria-en
+# Deepgram TTS playback speed (range: 0.5–1.5, default: 1.3).
+# Values below 1.0 slow down speech; values above 1.0 speed it up.
+#DEEPGRAM_TTS_SPEED=1.3
 # API key for Cartesia Sonic-3 TTS. Required when DISCOCLAW_TTS_PROVIDER=cartesia.
 #CARTESIA_API_KEY=

package/README.md

@@ -221,6 +221,18 @@ Full step-by-step guide: [docs/discord-bot-setup.md](docs/discord-bot-setup.md)
    npm install -g discoclaw
    ```
 
+   > **Fedora 43+ / GCC 14+ — `@discordjs/opus` build failure**
+   >
+   > GCC 14 promotes `-Wincompatible-pointer-types` to a hard error by default. The upstream opus C source triggers this, causing `npm install` to fail with an error like:
+   > ```
+   > error: incompatible pointer types passing ...
+   > ```
+   > **Workaround** — set the flag before installing:
+   > ```bash
+   > CFLAGS="-Wno-error=incompatible-pointer-types" npm install -g discoclaw
+   > ```
+   > This is a known upstream issue in the `@discordjs/opus` native addon. It only requires the flag override at install time; runtime behavior is unaffected.
+
 2. **Run the interactive setup wizard** (creates `.env` and scaffolds your workspace):
    ```bash
    discoclaw init

package/dist/config.js

@@ -158,6 +158,8 @@ export function parseConfig(env) {
     const discordActionsDefer = parseBoolean(env, 'DISCOCLAW_DISCORD_ACTIONS_DEFER', true);
     const discordActionsImagegen = parseBoolean(env, 'DISCOCLAW_DISCORD_ACTIONS_IMAGEGEN', false);
     const discordActionsVoice = parseBoolean(env, 'DISCOCLAW_DISCORD_ACTIONS_VOICE', false);
+    const discordActionsSpawn = parseBoolean(env, 'DISCOCLAW_DISCORD_ACTIONS_SPAWN', true);
+    const spawnMaxConcurrent = parsePositiveInt(env, 'DISCOCLAW_DISCORD_ACTIONS_SPAWN_MAX_CONCURRENT', 4);
     const deferMaxDelaySeconds = parsePositiveNumber(env, 'DISCOCLAW_DISCORD_ACTIONS_DEFER_MAX_DELAY_SECONDS', DEFAULT_DISCORD_ACTIONS_DEFER_MAX_DELAY_SECONDS);
     const deferMaxConcurrent = parsePositiveInt(env, 'DISCOCLAW_DISCORD_ACTIONS_DEFER_MAX_CONCURRENT', DEFAULT_DISCORD_ACTIONS_DEFER_MAX_CONCURRENT);
     if (!discordActionsEnabled) {
@@ -176,6 +178,7 @@ export function parseConfig(env) {
             { name: 'DISCOCLAW_DISCORD_ACTIONS_DEFER', enabled: discordActionsDefer },
             { name: 'DISCOCLAW_DISCORD_ACTIONS_IMAGEGEN', enabled: discordActionsImagegen },
             { name: 'DISCOCLAW_DISCORD_ACTIONS_VOICE', enabled: discordActionsVoice },
+            { name: 'DISCOCLAW_DISCORD_ACTIONS_SPAWN', enabled: discordActionsSpawn },
         ]
             .filter((entry) => (env[entry.name] ?? '').trim().length > 0 && entry.enabled)
             .map((entry) => entry.name);
@@ -243,6 +246,16 @@ export function parseConfig(env) {
     const deepgramApiKey = parseTrimmedString(env, 'DEEPGRAM_API_KEY');
     const deepgramSttModel = parseTrimmedString(env, 'DEEPGRAM_STT_MODEL') ?? 'nova-3-general';
     const deepgramTtsVoice = parseTrimmedString(env, 'DEEPGRAM_TTS_VOICE') ?? 'aura-2-asteria-en';
+    const deepgramTtsSpeed = (() => {
+        const raw = parseTrimmedString(env, 'DEEPGRAM_TTS_SPEED');
+        if (raw == null)
+            return 1.3;
+        const n = parseFloat(raw);
+        if (!Number.isFinite(n) || n < 0.5 || n > 1.5) {
+            throw new Error(`DEEPGRAM_TTS_SPEED must be a number between 0.5 and 1.5, got "${raw}"`);
+        }
+        return n;
+    })();
     const cartesiaApiKey = parseTrimmedString(env, 'CARTESIA_API_KEY');
     const voiceModelRaw = parseTrimmedString(env, 'DISCOCLAW_VOICE_MODEL');
     const voiceSystemPrompt = (() => {
@@ -351,8 +364,10 @@ export function parseConfig(env) {
             discordActionsDefer,
             discordActionsImagegen,
             discordActionsVoice,
+            discordActionsSpawn,
             deferMaxDelaySeconds,
             deferMaxConcurrent,
+            spawnMaxConcurrent,
             messageHistoryBudget: parseNonNegativeInt(env, 'DISCOCLAW_MESSAGE_HISTORY_BUDGET', 3000),
             summaryEnabled: parseBoolean(env, 'DISCOCLAW_SUMMARY_ENABLED', true),
             summaryModel: parseTrimmedString(env, 'DISCOCLAW_SUMMARY_MODEL') ?? fastModel,
@@ -398,6 +413,7 @@ export function parseConfig(env) {
             deepgramApiKey,
             deepgramSttModel,
             deepgramTtsVoice,
+            deepgramTtsSpeed,
             cartesiaApiKey,
             forgeDrafterRuntime,
             forgeAuditorRuntime,

package/dist/cron/auto-tag.js

@@ -59,7 +59,7 @@ export async function autoTagCron(runtime, name, prompt, availableTags, opts) {
 // Model tier classification
 // ---------------------------------------------------------------------------
 /**
- * Classify whether a cron job needs capable-tier or can run on fast.
+ * Classify whether a cron job needs deep/capable-tier or can run on fast.
  *
  * Two-step logic:
  * 1. Cadence default: frequent/hourly (>1x/day) → fast immediately (cost optimization).
@@ -70,10 +70,11 @@ export async function classifyCronModel(runtime, name, prompt, cadence, opts) {
     if (cadence === 'frequent' || cadence === 'hourly') {
         return 'fast';
     }
-    const classifyPrompt = `Does this scheduled task require advanced reasoning (complex analysis, ` +
-        `multi-step planning, nuanced writing) or can it be handled with basic ` +
-        `capabilities (simple lookups, templated responses, data formatting)?\n\n` +
-        `Reply with ONLY one word: "capable" or "fast"\n\n` +
+    const classifyPrompt = `Does this scheduled task require heavy reasoning (complex analysis, ` +
+        `multi-step planning, deep code review), standard advanced capabilities ` +
+        `(nuanced writing, moderate analysis), or basic capabilities ` +
+        `(simple lookups, templated responses, data formatting)?\n\n` +
+        `Reply with ONLY one word: "deep", "capable", or "fast"\n\n` +
         `Job name: ${name}\n` +
         `Instruction: ${prompt.slice(0, 500)}`;
     let finalText = '';
@@ -96,5 +97,7 @@ export async function classifyCronModel(runtime, name, prompt, cadence, opts) {
         }
     }
     const output = (finalText || deltaText).trim().toLowerCase();
-    return output === 'capable' ? 'capable' : 'fast';
+    if (output === 'capable' || output === 'deep')
+        return output;
+    return 'fast';
 }

package/dist/cron/cron-sync-coordinator.test.js

@@ -9,6 +9,7 @@ vi.mock('./cron-sync.js', () => ({
         tagsApplied: 1,
         namesUpdated: 0,
         statusMessagesUpdated: 1,
+        promptMessagesCreated: 0,
         orphansDetected: 0,
     })),
 }));
@@ -40,7 +41,7 @@ describe('CronSyncCoordinator', () => {
         vi.resetAllMocks();
         mockReload.mockResolvedValue(2);
         mockRunCronSync.mockResolvedValue({
-            tagsApplied: 1, namesUpdated: 0, statusMessagesUpdated: 1, orphansDetected: 0,
+            tagsApplied: 1, namesUpdated: 0, statusMessagesUpdated: 1, promptMessagesCreated: 0, orphansDetected: 0,
         });
     });
     it('reloads tag map before sync when tagMapPath is set', async () => {
@@ -69,7 +70,7 @@ describe('CronSyncCoordinator', () => {
     it('coalesced concurrent sync returns null', async () => {
         let resolveSync;
         mockRunCronSync.mockImplementation(() => new Promise((resolve) => {
-            resolveSync = () => resolve({ tagsApplied: 0, namesUpdated: 0, statusMessagesUpdated: 0, orphansDetected: 0 });
+            resolveSync = () => resolve({ tagsApplied: 0, namesUpdated: 0, statusMessagesUpdated: 0, promptMessagesCreated: 0, orphansDetected: 0 });
         }));
         const coordinator = new CronSyncCoordinator(makeOpts());
         const first = coordinator.sync();
@@ -85,10 +86,10 @@ describe('CronSyncCoordinator', () => {
         mockRunCronSync.mockImplementation(() => new Promise((resolve) => {
             callCount++;
             if (callCount === 1) {
-                resolveSync = () => resolve({ tagsApplied: 0, namesUpdated: 0, statusMessagesUpdated: 0, orphansDetected: 0 });
+                resolveSync = () => resolve({ tagsApplied: 0, namesUpdated: 0, statusMessagesUpdated: 0, promptMessagesCreated: 0, orphansDetected: 0 });
             }
             else {
-                resolve({ tagsApplied: 0, namesUpdated: 0, statusMessagesUpdated: 0, orphansDetected: 0 });
+                resolve({ tagsApplied: 0, namesUpdated: 0, statusMessagesUpdated: 0, promptMessagesCreated: 0, orphansDetected: 0 });
             }
         }));
         const coordinator = new CronSyncCoordinator(makeOpts());

package/dist/cron/cron-sync.js

@@ -1,3 +1,4 @@
+import { EmbedBuilder } from 'discord.js';
 import { CADENCE_TAGS } from './run-stats.js';
 import { detectCadence } from './cadence.js';
 import { autoTagCron, classifyCronModel } from './auto-tag.js';
@@ -24,13 +25,14 @@ export async function runCronSync(opts) {
     const forum = await resolveForumChannel(client, forumId);
     if (!forum) {
         log?.warn({ forumId }, 'cron-sync: forum not found');
-        return { tagsApplied: 0, namesUpdated: 0, statusMessagesUpdated: 0, orphansDetected: 0 };
+        return { tagsApplied: 0, namesUpdated: 0, statusMessagesUpdated: 0, promptMessagesCreated: 0, orphansDetected: 0 };
     }
     const tagMap = opts.tagMap;
     const purposeTags = purposeTagNames(tagMap);
     let tagsApplied = 0;
     let namesUpdated = 0;
     let statusMessagesUpdated = 0;
+    let promptMessagesCreated = 0;
     let orphansDetected = 0;
     const asEditableCronThread = (value) => {
         if (!value || typeof value !== 'object')
@@ -172,6 +174,51 @@ export async function runCronSync(opts) {
         }
         await sleep(throttleMs);
     }
+    // Phase 3.5: Prompt message backfill.
+    for (const job of jobs) {
+        const fullJob = scheduler.getJob(job.id);
+        if (!fullJob?.cronId)
+            continue;
+        const record = statsStore.getRecord(fullJob.cronId);
+        if (!record?.prompt || record.promptMessageId)
+            continue;
+        try {
+            let thread = null;
+            const cached = client.channels.cache.get(fullJob.threadId);
+            if (cached && cached.isThread()) {
+                thread = cached;
+            }
+            else {
+                try {
+                    const fetched = await client.channels.fetch(fullJob.threadId);
+                    if (fetched && fetched.isThread())
+                        thread = fetched;
+                }
+                catch {
+                    // Thread may have been deleted.
+                }
+            }
+            if (!thread)
+                continue;
+            const embed = new EmbedBuilder()
+                .setTitle('\uD83D\uDCCB Cron Prompt')
+                .setDescription(record.prompt.slice(0, 4096))
+                .setColor(0x5865F2);
+            const msg = await thread.send({ embeds: [embed], allowedMentions: { parse: [] } });
+            try {
+                await msg.pin();
+            }
+            catch {
+                // Non-fatal if pin fails.
+            }
+            await statsStore.upsertRecord(record.cronId, record.threadId, { promptMessageId: msg.id });
+            promptMessagesCreated++;
+        }
+        catch (err) {
+            log?.warn({ err, cronId: fullJob.cronId }, 'cron-sync:phase3.5 prompt message failed');
+        }
+        await sleep(throttleMs);
+    }
     // Phase 4: Orphan detection (non-destructive, log only).
     for (const thread of threads.values()) {
         const editableThread = asEditableCronThread(thread);
@@ -184,6 +231,6 @@ export async function runCronSync(opts) {
             log?.warn({ threadId: editableThread.id, name: editableThread.name }, 'cron-sync:phase4 orphan thread (no registered job)');
         }
     }
-    log?.info({ tagsApplied, namesUpdated, statusMessagesUpdated, orphansDetected }, 'cron-sync: complete');
-    return { tagsApplied, namesUpdated, statusMessagesUpdated, orphansDetected };
+    log?.info({ tagsApplied, namesUpdated, statusMessagesUpdated, promptMessagesCreated, orphansDetected }, 'cron-sync: complete');
+    return { tagsApplied, namesUpdated, statusMessagesUpdated, promptMessagesCreated, orphansDetected };
 }

package/dist/cron/cron-tag-map-watcher.test.js

@@ -12,7 +12,7 @@ function mockLog() {
 }
 function makeCoordinator() {
     return {
-        sync: vi.fn(async () => ({ tagsApplied: 0, namesUpdated: 0, statusMessagesUpdated: 0, orphansDetected: 0 })),
+        sync: vi.fn(async () => ({ tagsApplied: 0, namesUpdated: 0, statusMessagesUpdated: 0, promptMessagesCreated: 0, orphansDetected: 0 })),
     };
 }
 describe('startCronTagMapWatcher', () => {

package/dist/cron/executor.js

@@ -1,5 +1,6 @@
 import { acquireCronLock, releaseCronLock } from './job-lock.js';
 import { resolveChannel } from '../discord/action-utils.js';
+import { DiscordTransportClient } from '../discord/transport-client.js';
 import * as discordActions from '../discord/actions.js';
 import { sendChunks, appendUnavailableActionTypesNotice, appendParseFailureNotice } from '../discord/output-common.js';
 import { buildPromptPreamble, loadWorkspacePaFiles, inlineContextFiles, resolveEffectiveTools } from '../discord/prompt-common.js';
@@ -253,6 +254,7 @@ export async function executeCronJob(job, ctx) {
                     channelId: targetChannel.id,
                     messageId: '',
                     deferScheduler: ctx.deferScheduler,
+                    transport: new DiscordTransportClient(guild, ctx.client),
                     confirmation: {
                         mode: 'automated',
                     },
@@ -270,11 +272,8 @@ export async function executeCronJob(job, ctx) {
                     metrics.recordActionResult(result.ok);
                     ctx.log?.info({ flow: 'cron', jobId: job.id, ok: result.ok }, 'obs.action.result');
                 }
-                const displayLines = discordActions.buildDisplayResultLines(actions, results);
                 const anyActionSucceeded = results.some((r) => r.ok);
-                processedText = displayLines.length > 0
-                    ? cleanText.trimEnd() + '\n\n' + displayLines.join('\n')
-                    : cleanText.trimEnd();
+                processedText = discordActions.appendActionResults(cleanText.trimEnd(), actions, results);
                 // When all display lines were suppressed and there's no prose, skip posting.
                 if (!processedText.trim() && anyActionSucceeded && strippedUnrecognizedTypes.length === 0 && parseFailuresCount === 0) {
                     ctx.log?.info({ jobId: job.id }, 'cron:reply suppressed (actions-only, no display text)');

package/dist/cron/run-stats.js

@@ -5,7 +5,7 @@ import crypto from 'node:crypto';
 // Types
 // ---------------------------------------------------------------------------
 export const CADENCE_TAGS = ['yearly', 'frequent', 'hourly', 'daily', 'weekly', 'monthly'];
-export const CURRENT_VERSION = 7;
+export const CURRENT_VERSION = 8;
 // ---------------------------------------------------------------------------
 // Stable Cron ID generation
 // ---------------------------------------------------------------------------
@@ -313,5 +313,9 @@ export async function loadRunStats(filePath) {
     if (store.version === 6) {
         store.version = 7;
     }
+    // Migrate v7 → v8: no-op — new field (promptMessageId) is optional and defaults to absent.
+    if (store.version === 7) {
+        store.version = 8;
+    }
     return new CronRunStats(store, filePath);
 }

package/dist/cron/run-stats.test.js

@@ -38,7 +38,7 @@ describe('CronRunStats', () => {
     it('creates empty store on missing file', async () => {
         const stats = await loadRunStats(statsPath);
         const store = stats.getStore();
-        expect(store.version).toBe(7);
+        expect(store.version).toBe(8);
         expect(Object.keys(store.jobs)).toHaveLength(0);
     });
     it('upserts and retrieves records by cronId', async () => {
@@ -244,7 +244,7 @@ describe('CronRunStats', () => {
 describe('emptyStore', () => {
     it('returns valid initial structure', () => {
         const store = emptyStore();
-        expect(store.version).toBe(7);
+        expect(store.version).toBe(8);
         expect(store.updatedAt).toBeGreaterThan(0);
         expect(Object.keys(store.jobs)).toHaveLength(0);
     });
@@ -271,7 +271,7 @@ describe('loadRunStats version migration', () => {
         };
         await fs.writeFile(statsPath, JSON.stringify(v3Store), 'utf-8');
         const stats = await loadRunStats(statsPath);
-        expect(stats.getStore().version).toBe(7);
+        expect(stats.getStore().version).toBe(8);
         const rec = stats.getRecord('cron-migrated');
         expect(rec).toBeDefined();
         expect(rec.cronId).toBe('cron-migrated');
@@ -301,7 +301,7 @@ describe('loadRunStats version migration', () => {
         };
         await fs.writeFile(statsPath, JSON.stringify(v4Store), 'utf-8');
         const stats = await loadRunStats(statsPath);
-        expect(stats.getStore().version).toBe(7);
+        expect(stats.getStore().version).toBe(8);
         const rec = stats.getRecord('cron-v4');
         expect(rec).toBeDefined();
         expect(rec.cronId).toBe('cron-v4');
@@ -331,7 +331,7 @@ describe('loadRunStats version migration', () => {
         };
         await fs.writeFile(statsPath, JSON.stringify(v5Store), 'utf-8');
         const stats = await loadRunStats(statsPath);
-        expect(stats.getStore().version).toBe(7);
+        expect(stats.getStore().version).toBe(8);
         const rec = stats.getRecord('cron-v5');
         expect(rec).toBeDefined();
         expect(rec.cronId).toBe('cron-v5');
@@ -367,7 +367,7 @@ describe('loadRunStats version migration', () => {
         };
         await fs.writeFile(statsPath, JSON.stringify(v6Store), 'utf-8');
         const stats = await loadRunStats(statsPath);
-        expect(stats.getStore().version).toBe(7);
+        expect(stats.getStore().version).toBe(8);
         const rec = stats.getRecord('cron-v6');
         expect(rec).toBeDefined();
         expect(rec.cronId).toBe('cron-v6');

package/dist/discord/action-categories.js

@@ -32,6 +32,8 @@ export const QUERY_ACTION_TYPES = new Set([
     'forgeStatus',
     // Voice
     'voiceStatus',
+    // Spawn
+    'spawnAgent',
 ]);
 export function hasQueryAction(actionTypes) {
     return actionTypes.some((t) => QUERY_ACTION_TYPES.has(t));

package/dist/discord/actions-config.js

@@ -297,7 +297,7 @@ export function configActionsPromptSection() {
 <discord-action>{"type":"modelSet","role":"fast","model":"haiku"}</discord-action>
 \`\`\`
 - \`role\` (required): One of \`chat\`, \`fast\`, \`forge-drafter\`, \`forge-auditor\`, \`summary\`, \`cron\`, \`cron-exec\`, \`voice\`.
-- \`model\` (required): Model tier (\`fast\`, \`capable\`), concrete model name (\`haiku\`, \`sonnet\`, \`opus\`), runtime name (\`openrouter\`, \`gemini\` — for \`chat\` role, swaps the active runtime adapter), or \`default\` (for cron-exec only, to revert to the env-configured default (Sonnet by default)).
+- \`model\` (required): Model tier (\`fast\`, \`capable\`, \`deep\`), concrete model name (\`haiku\`, \`sonnet\`, \`opus\`), runtime name (\`openrouter\`, \`gemini\` — for \`chat\` role, swaps the active runtime adapter), or \`default\` (for cron-exec only, to revert to the env-configured default (Sonnet by default)).
 
 **Roles:**
 | Role | What it controls |