@vellumai/assistant 0.4.56 → 0.4.57

package/src/config/bundled-skills/google-calendar/TOOLS.json

@@ -38,7 +38,7 @@
             "enum": ["startTime", "updated"],
             "description": "Sort order (default \"startTime\"). Only valid when single_events is true."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           },
@@ -67,7 +67,7 @@
             "type": "string",
             "description": "Calendar ID (default \"primary\")"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           },
@@ -128,7 +128,7 @@
             "type": "number",
             "description": "Confidence score (0-1) for this action"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           },
@@ -169,7 +169,7 @@
             "type": "string",
             "description": "IANA timezone for interpreting results (e.g. \"America/New_York\")"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           },
@@ -208,7 +208,7 @@
             "type": "number",
             "description": "Confidence score (0-1) for this action"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           },

package/src/config/bundled-skills/google-calendar/types.ts

@@ -1,4 +1,4 @@
-/** Event time — either a dateTime with timezone or a date for all-day events. */
+/** Event time - either a dateTime with timezone or a date for all-day events. */
 export interface EventDateTime {
   dateTime?: string;
   date?: string;

package/src/config/bundled-skills/heartbeat/SKILL.md

@@ -0,0 +1,43 @@
+---
+name: heartbeat
+description: Configure periodic background checklist runs
+compatibility: "Designed for Vellum personal assistants"
+metadata:
+  emoji: "\U0001F493"
+  vellum:
+    display-name: "Heartbeat"
+    activation-hints:
+      - "Set up a heartbeat, periodic checklist, background health check, or recurring background task"
+    avoid-when:
+      - "One-off or recurring schedules with specific payloads - use the schedule skill instead"
+---
+
+The heartbeat feature runs your `HEARTBEAT.md` checklist periodically in a background conversation. Each run, the assistant works through the checklist and flags anything that needs attention.
+
+## Setup
+
+Edit `config.json` using `file_edit`:
+
+1. **Enable heartbeat**: Set `heartbeat.enabled` to `true`.
+2. **Set interval**: Set `heartbeat.intervalMs` (milliseconds between runs, default: 3600000 = 1 hour).
+3. **Optional active hours**: Set `heartbeat.activeHoursStart` and `heartbeat.activeHoursEnd` (0-23) to restrict runs to certain hours. Both must be set together.
+
+Example config.json heartbeat section:
+```json
+{
+  "heartbeat": {
+    "enabled": true,
+    "intervalMs": 1800000,
+    "activeHoursStart": 8,
+    "activeHoursEnd": 22
+  }
+}
+```
+
+Then edit `HEARTBEAT.md` with the checklist items. The assistant will work through this file each heartbeat run.
+
+## Notes
+
+- Toggling `heartbeat.enabled` requires an assistant restart to take effect.
+- Changes to `HEARTBEAT.md` take effect on the next heartbeat run (no restart needed).
+- The heartbeat runs in a separate background conversation, not the user's active chat.

package/src/config/bundled-skills/image-studio/SKILL.md

@@ -23,12 +23,12 @@ You are an image generation assistant. When the user asks you to create or edit
 
 ## Models
 
-- `gemini-2.5-flash-image` (default) — fast, good quality
-- `gemini-3-pro-image` — higher quality, slower
+- `gemini-3.1-flash-image-preview` (default) - Nano Banana 2, fast, good quality
+- `gemini-3-pro-image-preview` - Nano Banana Pro, higher quality, slower
 
 ## Tips
 
 - Be descriptive in your prompts for better results. Include details about style, composition, lighting, and mood.
 - When editing images, clearly describe what changes you want made to the source image.
 - Use the `variants` parameter (1-4) to generate multiple options and pick the best one.
-- If no Gemini API key is configured, the tool will return an error — ask the user to set one up.
+- If no Gemini API key is configured, the tool will return an error - ask the user to set one up.

package/src/config/bundled-skills/image-studio/TOOLS.json

@@ -28,8 +28,7 @@
           "model": {
             "type": "string",
             "enum": [
-              "gemini-2.5-flash-image",
-              "gemini-3-pro-image",
+              "gemini-3.1-flash-image-preview",
               "gemini-3-pro-image-preview"
             ],
             "description": "Which model to use for generation. If omitted, uses the user's configured preference."
@@ -38,7 +37,7 @@
             "type": "number",
             "description": "Number of image variants to generate (1-4, default: 1)"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }

package/src/config/bundled-skills/image-studio/tools/media-generate-image.ts

@@ -41,7 +41,7 @@ function isAttachmentAccessible(
   if (hasStandard) {
     return true;
   }
-  // All sources are private — visible only if the caller is in one of those conversations
+  // All sources are private - visible only if the caller is in one of those conversations
   return sources.some((s) =>
     isAttachmentVisible(
       { conversationId: s.conversationId, isPrivate: true },
@@ -55,14 +55,13 @@ export async function run(
   context: ToolContext,
 ): Promise<ToolExecutionResult> {
   const config = getConfig();
-  const apiKey = await getProviderKeyAsync("gemini");
+  const imageGenMode = config.services["image-generation"].mode;
 
-  // Resolve credentials: prefer direct API key, fall back to managed proxy
+  // Resolve credentials strictly based on mode — no cross-mode fallbacks
   let credentials: ImageGenCredentials | undefined;
-  if (apiKey) {
-    credentials = { type: "direct", apiKey };
-  } else {
-    const managedBaseUrl = await buildManagedBaseUrl("vertex");
+
+  if (imageGenMode === "managed") {
+    const managedBaseUrl = await buildManagedBaseUrl("gemini");
     if (managedBaseUrl) {
       const ctx = await resolveManagedProxyContext();
       credentials = {
@@ -71,14 +70,19 @@ export async function run(
         baseUrl: managedBaseUrl,
       };
     }
+  } else {
+    const apiKey = await getProviderKeyAsync("gemini");
+    if (apiKey) {
+      credentials = { type: "direct", apiKey };
+    }
   }
 
   if (!credentials) {
-    return {
-      content:
-        "No Gemini API key configured. Please set your Gemini API key to use image generation.",
-      isError: true,
-    };
+    const hint =
+      imageGenMode === "managed"
+        ? "Managed proxy is not available. Please log in to Vellum or switch to Your Own mode."
+        : "No Gemini API key configured. Please set your Gemini API key in Settings > Models & Services.";
+    return { content: hint, isError: true };
   }
 
   const prompt = input.prompt as string;

package/src/config/bundled-skills/media-processing/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: media-processing
-description: "Ingest and process media files (video, audio, image) through a 3-phase pipeline: preprocess, map (Gemini), and reduce (Claude)"
+description: "Ingest and process media files (video, audio, image)"
 compatibility: "Designed for Vellum personal assistants"
 metadata:
   emoji: "🎬"
@@ -14,11 +14,11 @@ Ingest and track processing of media files (video, audio, images) through a conf
 
 The processing pipeline follows a sequential 3-phase flow:
 
-1. **Ingest** (`ingest_media`) — Register a media file, detect MIME type, extract duration, deduplicate by content hash.
-2. **Preprocess** (`extract_keyframes`) — Detect dead time, segment the video into windows, extract downscaled keyframes, build a subject registry, and write a pipeline manifest.
-3. **Map** (`analyze_keyframes`) — Send each segment's frames to Gemini 2.5 Flash with assistant-provided extraction instructions and a JSON Schema for guaranteed structured output. Supports concurrency pooling, cost tracking, resumability, and automatic retries.
-4. **Reduce / Query** (`query_media`) — Send all map output to Claude for intelligent analysis and Q&A. Supports arbitrary natural language queries about video content.
-5. **Clip** (`generate_clip`) — Extract video clips around specific moments.
+1. **Ingest** (`ingest_media`) - Register a media file, detect MIME type, extract duration, deduplicate by content hash.
+2. **Preprocess** (`extract_keyframes`) - Detect dead time, segment the video into windows, extract downscaled keyframes, build a subject registry, and write a pipeline manifest.
+3. **Map** (`analyze_keyframes`) - Send each segment's frames to Gemini 2.5 Flash with assistant-provided extraction instructions and a JSON Schema for guaranteed structured output. Supports concurrency pooling, cost tracking, resumability, and automatic retries.
+4. **Reduce / Query** (`query_media`) - Send all map output to Claude for intelligent analysis and Q&A. Supports arbitrary natural language queries about video content.
+5. **Clip** (`generate_clip`) - Extract video clips around specific moments.
 
 The processing pipeline service (`services/processing-pipeline.ts`) orchestrates phases 2-4 automatically with retries, resumability, and cancellation support.
 
@@ -38,35 +38,35 @@ Preprocess a video asset: detect dead time via mpdecimate, segment the video int
 
 Parameters:
 
-- `asset_id` (required) — ID of the media asset.
-- `interval_seconds` — Interval between keyframes (default: 1s). Use 0.5s for sports/action content where frame density matters.
-- `segment_duration` — Duration of each segment window (default: 15s).
-- `dead_time_threshold` — Sensitivity for dead-time detection (default: 0.02).
-- `section_config` — Path to a JSON file with manual section boundaries.
-- `detect_dead_time` — Whether to detect and skip dead time (default: false). Dead-time detection can be too aggressive for continuous action video like sports — it may incorrectly skip live play. Enable only for content with clear idle periods (e.g., lectures, surveillance footage).
-- `short_edge` — Short edge resolution for downscaled frames in pixels (default: 480).
-- `include_audio` — Whether to extract and transcribe audio for each segment (default: false). When enabled, each segment's audio is transcribed and stored alongside visual frames.
-- `transcription_mode` — Transcription backend: `'api'` (OpenAI Whisper cloud) or `'local'` (whisper.cpp on-device). Default: `'local'`. The `'api'` mode requires an OpenAI API key configured in settings.
+- `asset_id` (required) - ID of the media asset.
+- `interval_seconds` - Interval between keyframes (default: 1s). Use 0.5s for sports/action content where frame density matters.
+- `segment_duration` - Duration of each segment window (default: 15s).
+- `dead_time_threshold` - Sensitivity for dead-time detection (default: 0.02).
+- `section_config` - Path to a JSON file with manual section boundaries.
+- `detect_dead_time` - Whether to detect and skip dead time (default: false). Dead-time detection can be too aggressive for continuous action video like sports - it may incorrectly skip live play. Enable only for content with clear idle periods (e.g., lectures, surveillance footage).
+- `short_edge` - Short edge resolution for downscaled frames in pixels (default: 480).
+- `include_audio` - Whether to extract and transcribe audio for each segment (default: false). When enabled, each segment's audio is transcribed and stored alongside visual frames.
+- `transcription_mode` - Transcription backend: `'api'` (OpenAI Whisper cloud) or `'local'` (whisper.cpp on-device). Default: `'local'`. The `'api'` mode requires an OpenAI API key configured in settings.
 
 ### analyze_keyframes
 
 Map video segments through Gemini's structured output API. Supports two modes:
 
-- **`keyframes`** (default) — Reads frames from the preprocess manifest, sends each segment's images to Gemini. Requires `extract_keyframes` to be run first. Best for longer videos (> 1 hour) or when you need fine-grained control over frame selection (interval, segment duration, dead-time skipping).
-- **`direct_video`** — Uploads the video file directly to Gemini's Files API. Gemini sees actual motion and temporal context instead of static frames. Best for shorter videos (< 1 hour) where temporal context matters (detecting actions, transitions, motion patterns). Has a 2 GB file size limit. Does not require `extract_keyframes` preprocessing.
+- **`keyframes`** (default) - Reads frames from the preprocess manifest, sends each segment's images to Gemini. Requires `extract_keyframes` to be run first. Best for longer videos (> 1 hour) or when you need fine-grained control over frame selection (interval, segment duration, dead-time skipping).
+- **`direct_video`** - Uploads the video file directly to Gemini's Files API. Gemini sees actual motion and temporal context instead of static frames. Best for shorter videos (< 1 hour) where temporal context matters (detecting actions, transitions, motion patterns). Has a 2 GB file size limit. Does not require `extract_keyframes` preprocessing.
 
 Both modes produce the same `MapOutput` format, so `query_media` works identically regardless of which mode was used.
 
 Parameters:
 
-- `asset_id` (required) — ID of the media asset.
-- `system_prompt` (required) — Extraction instructions for Gemini.
-- `output_schema` (required) — JSON Schema for structured output.
-- `mode` — Analysis mode: `'keyframes'` (default) or `'direct_video'`.
-- `context` — Additional context to include in the prompt.
-- `model` — Gemini model to use (default: `gemini-2.5-flash`).
-- `concurrency` — Maximum concurrent API requests (default: 10, keyframes mode only).
-- `max_retries` — Retry attempts per segment on failure (default: 3).
+- `asset_id` (required) - ID of the media asset.
+- `system_prompt` (required) - Extraction instructions for Gemini.
+- `output_schema` (required) - JSON Schema for structured output.
+- `mode` - Analysis mode: `'keyframes'` (default) or `'direct_video'`.
+- `context` - Additional context to include in the prompt.
+- `model` - Gemini model to use (default: `gemini-2.5-flash`).
+- `concurrency` - Maximum concurrent API requests (default: 10, keyframes mode only).
+- `max_retries` - Retry attempts per segment on failure (default: 3).
 
 ### query_media
 
@@ -74,10 +74,10 @@ Query video analysis data using natural language. Sends map output (from analyze
 
 Parameters:
 
-- `asset_id` (required) — ID of the media asset.
-- `query` (required) — Natural language query about the video data.
-- `system_prompt` — Optional system prompt for Claude.
-- `model` — LLM model to use (default: `claude-sonnet-4-6`).
+- `asset_id` (required) - ID of the media asset.
+- `query` (required) - Natural language query about the video data.
+- `system_prompt` - Optional system prompt for Claude.
+- `model` - LLM model to use (default: `claude-sonnet-4-6`).
 
 ### generate_clip
 
@@ -135,7 +135,7 @@ Transcription modes:
 - **`local`** (default): Uses whisper.cpp for on-device transcription. Requires `whisper-cpp` to be installed (`brew install whisper-cpp`). No API costs, but slower.
 - **`api`**: Uses the OpenAI Whisper API for cloud-based transcription. Faster and more accurate, but requires an OpenAI API key and incurs per-minute costs.
 
-The audio transcription degrades gracefully — if transcription fails for a segment (missing tools, no audio track, API errors), the segment proceeds with visual-only analysis.
+The audio transcription degrades gracefully - if transcription fails for a segment (missing tools, no audio track, API errors), the segment proceeds with visual-only analysis.
 
 ## Best Practices
 
@@ -195,13 +195,13 @@ Be specific and factual. Describe what you see, not what you infer happened betw
 
 ### Clip Delivery
 
-The `generate_clip` tool automatically opens clips in the user's default video player after extraction (handled internally — do **not** run `open` via `host_bash`). Clips are saved persistently in the asset's pipeline directory (`pipeline/<assetId>/clips/`), falling back to a temp directory when the source location is read-only. Each clip gets a unique filename so concurrent or repeated extractions at the same range never collide. The `clipPath` field in the tool response contains the absolute file path.
+The `generate_clip` tool automatically opens clips in the user's default video player after extraction (handled internally - do **not** run `open` via `host_bash`). Clips are saved persistently in the asset's pipeline directory (`pipeline/<assetId>/clips/`), falling back to a temp directory when the source location is read-only. Each clip gets a unique filename so concurrent or repeated extractions at the same range never collide. The `clipPath` field in the tool response contains the absolute file path.
 
-The tool handles high-bitrate and incompatible codec sources automatically — it tries stream copy first for speed, then falls back to H.264 re-encoding if needed. **Always use `generate_clip` rather than manual ffmpeg commands.**
+The tool handles high-bitrate and incompatible codec sources automatically - it tries stream copy first for speed, then falls back to H.264 re-encoding if needed. **Always use `generate_clip` rather than manual ffmpeg commands.**
 
 Always provide a descriptive `title` parameter (e.g. `"snow-dive-closeup"`, `"goal-celebration"`) so clips get meaningful filenames instead of timestamp-based names.
 
-## Known Limitations — Vision Analysis
+## Known Limitations - Vision Analysis
 
 Gemini performs well at **spatial/descriptive analysis** from static keyframes:
 
@@ -217,7 +217,7 @@ Gemini **hallucinates when asked to detect fast temporal events** from static fr
 - Fast transitions and split-second actions
 - Causality between frames (what "happened" vs. what's visible)
 
-The model is good at describing **what is there** but bad at detecting **what happened** from static frames. For content where temporal context matters, consider using `mode: 'direct_video'` which lets Gemini see actual motion. For keyframes mode, structure your map prompts and queries accordingly — ask the model to describe scenes, then use `query_media` (Claude) to reason about patterns and events across the descriptive data.
+The model is good at describing **what is there** but bad at detecting **what happened** from static frames. For content where temporal context matters, consider using `mode: 'direct_video'` which lets Gemini see actual motion. For keyframes mode, structure your map prompts and queries accordingly - ask the model to describe scenes, then use `query_media` (Claude) to reason about patterns and events across the descriptive data.
 
 ## Operator Runbook
 
@@ -225,10 +225,10 @@ The model is good at describing **what is there** but bad at detecting **what ha
 
 Use `media_status` to check the current state of any asset:
 
-- **registered** — Ingested but not yet processed.
-- **processing** — Pipeline is running.
-- **indexed** — All stages completed successfully.
-- **failed** — A stage failed. Check stage details for the error.
+- **registered** - Ingested but not yet processed.
+- **processing** - Pipeline is running.
+- **indexed** - All stages completed successfully.
+- **failed** - A stage failed. Check stage details for the error.
 
 The response includes per-stage progress (0-100%) so you can see exactly where processing stands.
 
@@ -244,11 +244,11 @@ Use `media_status` to check processing stages:
    - **map**: Gemini API key not configured, API rate limits, network errors.
    - **reduce**: No LLM provider configured, no map output exists.
 
-After fixing the root cause, re-run the failed stage. The pipeline is resumable — it picks up from where it left off.
+After fixing the root cause, re-run the failed stage. The pipeline is resumable - it picks up from where it left off.
 
 ### Cost Expectations
 
-The Map phase (Gemini) is the primary cost driver — it scales with video duration and keyframe interval. The Q&A phase (Claude) is negligible per query.
+The Map phase (Gemini) is the primary cost driver - it scales with video duration and keyframe interval. The Q&A phase (Claude) is negligible per query.
 
 ### Known Limitations
 

package/src/config/bundled-skills/media-processing/TOOLS.json

@@ -21,7 +21,7 @@
             "type": "object",
             "description": "Optional JSON metadata to attach to the asset (e.g., pipeline config, source info)"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -52,7 +52,7 @@
             "enum": ["registered", "processing", "indexed", "failed"],
             "description": "Filter assets by processing status"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -106,7 +106,7 @@
             "enum": ["api", "local"],
             "description": "Transcription backend: 'api' (OpenAI Whisper cloud) or 'local' (whisper.cpp). Default: 'local'."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -139,7 +139,7 @@
           },
           "output_schema": {
             "type": "object",
-            "description": "JSON Schema for structured output — Gemini will enforce this schema on the response"
+            "description": "JSON Schema for structured output - Gemini will enforce this schema on the response"
           },
           "context": {
             "type": "object",
@@ -159,7 +159,7 @@
             "minimum": 0,
             "description": "Maximum retry attempts per segment on failure. Default: 3"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -171,7 +171,7 @@
     },
     {
       "name": "query_media",
-      "description": "Query video analysis data using natural language. Sends map output (from analyze_keyframes) to Claude for intelligent analysis and Q&A. Supports arbitrary questions about video content — Claude reads the full structured analysis and answers based on the data.",
+      "description": "Query video analysis data using natural language. Sends map output (from analyze_keyframes) to Claude for intelligent analysis and Q&A. Supports arbitrary questions about video content - Claude reads the full structured analysis and answers based on the data.",
       "category": "media",
       "risk": "low",
       "input_schema": {
@@ -193,7 +193,7 @@
             "type": "string",
             "description": "LLM model to use for analysis. Default: 'claude-sonnet-4-6'"
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }
@@ -240,7 +240,7 @@
             "type": "string",
             "description": "Short descriptive title for the clip (e.g. 'snow-dive-closeup', 'goal-celebration'). Used as the filename. If omitted, falls back to timestamp-based naming."
           },
-          "reason": {
+          "activity": {
             "type": "string",
             "description": "Brief non-technical explanation of why this tool is being called"
           }

package/src/config/bundled-skills/media-processing/__tests__/concurrency-pool.test.ts

@@ -23,7 +23,7 @@ describe("ConcurrencyPool", () => {
     expect(thirdResolved).toBe(false);
     expect(pool.waitingCount).toBe(1);
 
-    // Release one slot — third should now resolve
+    // Release one slot - third should now resolve
     pool.release();
     await thirdPromise;
     expect(thirdResolved).toBe(true);
@@ -62,7 +62,7 @@ describe("ConcurrencyPool", () => {
   it("defaults to maxConcurrency of 10", async () => {
     const pool = new ConcurrencyPool();
 
-    // Acquire 10 slots — all should resolve immediately
+    // Acquire 10 slots - all should resolve immediately
     for (let i = 0; i < 10; i++) {
       await pool.acquire();
     }

package/src/config/bundled-skills/media-processing/__tests__/preprocess.test.ts

@@ -80,7 +80,7 @@ describe("buildDeadTimeRanges", () => {
   });
 
   it("filters out short ranges below minDuration", () => {
-    // Only 3 seconds of drops — below the 5s minimum
+    // Only 3 seconds of drops - below the 5s minimum
     const timestamps = [10, 10.5, 11, 11.5, 12, 12.5, 13];
     const ranges = buildDeadTimeRanges(timestamps, 1.0, 5.0);
     expect(ranges).toEqual([]);

package/src/config/bundled-skills/media-processing/services/gemini-map.ts

@@ -1,5 +1,5 @@
 /**
- * Gemini Map service — processes video segments through Gemini's structured
+ * Gemini Map service - processes video segments through Gemini's structured
  * output API for vision-based analysis with guaranteed valid JSON responses.
  *
  * Uses @google/genai SDK directly (not the GeminiProvider wrapper) to leverage
@@ -194,7 +194,7 @@ async function processSegmentWithRetry(
         },
       };
     } catch (err) {
-      // Handle Gemini safety blocks — not retryable
+      // Handle Gemini safety blocks - not retryable
       if (err instanceof ApiError) {
         const message = err.message ?? "";
         if (message.includes("SAFETY") || message.includes("safety")) {
@@ -208,7 +208,7 @@ async function processSegmentWithRetry(
           };
         }
 
-        // Non-retryable client errors (400, 401, 403, etc.) — fail immediately
+        // Non-retryable client errors (400, 401, 403, etc.) - fail immediately
         if (
           err.status !== undefined &&
           err.status < 500 &&
@@ -225,7 +225,7 @@ async function processSegmentWithRetry(
           };
         }
 
-        // 429 rate limits — retryable with backoff
+        // 429 rate limits - retryable with backoff
         if (err.status === 429 && attempt < maxRetries) {
           const delay = computeRetryDelay(attempt);
           onProgress?.(
@@ -332,7 +332,7 @@ export async function mapSegments(
         onProgress?.(`  Segment ${segment.id}: loaded from cache.\n`);
         return;
       } catch {
-        // Corrupted cache file — reprocess
+        // Corrupted cache file - reprocess
       }
     }
 

package/src/config/bundled-skills/media-processing/services/gemini-video.ts

@@ -1,5 +1,5 @@
 /**
- * Gemini Video service — uploads video directly to Gemini's Files API for
+ * Gemini Video service - uploads video directly to Gemini's Files API for
  * analysis, letting Gemini see actual motion and temporal context instead
  * of static keyframes.
  *
@@ -266,7 +266,7 @@ export async function analyzeVideoDirectly(
         await client.files.delete({ name: uploadedFileName });
         onProgress?.(`Cleaned up uploaded file from Gemini.\n`);
       } catch {
-        // Best-effort cleanup — don't fail the operation
+        // Best-effort cleanup - don't fail the operation
       }
     }
   }

package/src/config/bundled-skills/media-processing/services/preprocess.ts

@@ -319,7 +319,7 @@ async function extractDominantColors(framePath: string): Promise<string[]> {
   // Fallback: return empty if analysis fails
   if (result.exitCode !== 0) return [];
 
-  // Parse palette info from stderr — look for color hex values
+  // Parse palette info from stderr - look for color hex values
   const colors: string[] = [];
   const colorMatches = result.stderr.matchAll(/0x([0-9a-fA-F]{6})/g);
   for (const m of colorMatches) {
@@ -549,7 +549,7 @@ export async function preprocessForAsset(
     );
     if (rawSegments.length > 0 && totalFrames === 0) {
       throw new Error(
-        `All ${rawSegments.length} segment(s) failed frame extraction — zero usable frames produced.`,
+        `All ${rawSegments.length} segment(s) failed frame extraction - zero usable frames produced.`,
       );
     }
     onProgress?.(

package/src/config/bundled-skills/media-processing/services/processing-pipeline.ts

@@ -151,7 +151,7 @@ export async function runPipeline(
     resumedFrom = STAGE_ORDER[startIndex];
     onProgress?.(`Resuming pipeline from stage: ${resumedFrom}`);
   } else if (startIndex >= STAGE_ORDER.length) {
-    // All stages already completed — idempotent no-op
+    // All stages already completed - idempotent no-op
     onProgress?.("All pipeline stages already completed.");
     updateMediaAssetStatus(assetId, "indexed");
     return {
@@ -225,7 +225,7 @@ export async function runPipeline(
           `Stage ${stageName} failed (attempt ${attempt + 1}/${maxRetries + 1}): ${errorMsg}`,
         );
 
-        // Save partial progress — the stage handler should have already
+        // Save partial progress - the stage handler should have already
         // persisted any partial results before throwing
         updateProcessingStage(stageRecord.id, {
           status: attempt >= maxRetries ? "failed" : "running",

package/src/config/bundled-skills/media-processing/services/reduce.ts

@@ -1,5 +1,5 @@
 /**
- * Reduce service — sends Map output to Claude as text-only for analysis.
+ * Reduce service - sends Map output to Claude as text-only for analysis.
  *
  * Two modes:
  * - One-shot merge: assembles all Map results into a single document,
@@ -75,7 +75,7 @@ async function loadMapOutput(assetId: string): Promise<MapOutput> {
 
 /**
  * Format map output segments into a text document for Claude.
- * Strips image data — text only.
+ * Strips image data - text only.
  */
 function formatMapOutputAsText(mapOutput: MapOutput): string {
   const lines: string[] = [];
@@ -151,7 +151,7 @@ async function persistReduceCost(
     const raw = await readFile(costPath, "utf-8");
     existing = JSON.parse(raw) as ReduceCostData;
   } catch {
-    // First query — start fresh
+    // First query - start fresh
   }
 
   existing.entries.push({

package/src/config/bundled-skills/media-processing/tools/generate-clip.ts

@@ -1,5 +1,5 @@
 /**
- * Clip generation tool — extract a video segment from a media asset.
+ * Clip generation tool - extract a video segment from a media asset.
  *
  * Uses ffmpeg to cut a segment with configurable pre/post-roll padding,
  * then registers the resulting clip as an attachment for in-chat delivery.
@@ -199,7 +199,7 @@ export async function run(
     const result = await spawnWithTimeout(ffmpegArgs, FFMPEG_CLIP_TIMEOUT_MS);
 
     if (result.exitCode !== 0) {
-      // Stream copy failed — fall back to re-encoding (handles high-bitrate
+      // Stream copy failed - fall back to re-encoding (handles high-bitrate
       // sources, incompatible codecs, and missing keyframes at cut points)
       context.onOutput?.("Stream copy failed, re-encoding clip...\n");
       // Select codecs based on output format (WebM needs VP9/Opus, MP4/MOV use H.264/AAC)

package/src/config/bundled-skills/media-processing/tools/query-media-events.ts

@@ -1,5 +1,5 @@
 /**
- * Query media tool — sends natural language queries against video
+ * Query media tool - sends natural language queries against video
  * analysis data (map output) via Claude for intelligent answers.
  *
  * Replaces the old keyword-matching approach with an LLM-powered