@lalalic/markcut 2.2.7 → 2.2.9

package/skills/markcut/docs/templates/vlog/prompts/local-context.md

@@ -1,59 +0,0 @@
-# Prompt: Search Local Context (News, Weather, Events)
-
-> Fill every `{placeholder}`, then execute in your own context.
-> Runs after Phase 1 (media understanding) before storyboard writing.
-> Uses web search to find real facts about the vlog's location and date.
-
----
-
-You are researching local context for a vlog about "{theme}" at {location}.
-
-Date range: {date_range} (or approximate dates from media timestamps)
-Location description: {location} (from GPS reverse-geocode or user input)
-Vlog style: {style}
-
-## Task
-
-Search for the following and report real facts that could naturally weave into the narration:
-
-1. **Weather** — What was the weather like on those dates? High/low temps, conditions (sunny, rainy, cloudy), notable patterns (heatwave, storm, full moon).
-2. **Local events** — Any festivals, markets, sports events, holidays, or community gatherings happening at that location during the date range.
-3. **Seasonal context** — Seasonal highlights: fall colours, cherry blossoms, ski season opening, migration patterns, park hours.
-4. **Recent news** — Any notable local news near that location within the week before the vlog dates (park closures, trail openings, wildlife alerts, cultural events).
-
-Use multiple search queries for coverage. Prefer official sources (weather service, park website, local news).
-
-## Output format
-
-```
-## Local Context
-
-Weather:
-- <fact 1> (source: <url>)
-- <fact 2> (source: <url>)
-
-Events:
-- <fact> (source: <url>)
-
-Seasonal:
-- <fact> (source: <url>)
-
-News:
-- <fact> (source: <url>)
-
-## Amplification ideas (optional, for narrator)
-
-- <How a fact amplifies the story: "The fall colours festival meant the park was busier than usual, which explains the crowded trail clip">
-- <Another idea>
-```
-
-## Rules
-
-- Only include facts confirmed by at least one web source. If a search yields nothing useful for a category, write "None found."
-- Do not invent weather, events, or news.
-- The amplification ideas are suggestions only — the story-writer decides whether to use them.
-- Focus on facts that **connect** to what the media shows (weather matching the sky in photos, events explaining crowds, seasonal features matching the scenery).
-
-## After this step
-
-Pass the result to the storyboard phase (Phase 3). The story-writer agent or outline prompt should receive the local context as an additional input. Use at most **2 local-context references** in the final narration — enough to add color, not enough to overwhelm the personal story.

package/skills/markcut/docs/templates/vlog/prompts/outline.md

@@ -1,56 +0,0 @@
-# Prompt: Vlog Outline — Narrative Arc
-
-> Fill every `{placeholder}`, then execute in your own context.
-> Reads: `metadata.json` (per-clip captions, aes_score, GPS), project inputs.
-
----
-
-You are planning a {duration_min}-second {style} vlog: "{theme}".
-
-Creator: {creator_profile}
-Language: {language}
-
-Given the clip understanding below, plan the narrative arc.
-
-Clip understanding (from `markcut vision --label`):
-- Total clips: {total_clips}
-- Dominant mood: {dominant_mood}
-- Narrative summary: {narrative_summary}
-- Key subjects: {key_subjects}
-{clips_detail}
-
-Requirements:
-
-1. **Hook → Core → Vibe → Close arc**: First scene hooks the viewer with the best visual + intriguing line. Middle scenes are chronological story beats. Final scene provides closure.
-
-2. **Visual continuity**: Clips sharing same background/outfit/subject-state form one scene. No jumping A→B→A.
-
-3. **Drop rules**: Keep all clips unless aes_score < 0.4, near-duplicate, or off-topic. Max ~20% drops. Never below 5 clips.
-
-4. **GPS**: {gps_info} — if ≥2 GPS waypoints, include a route scene with map.
-
-Output format:
-
-```
-## Scene Plan
-
-### Route Scene (if applicable)
-- waypoints: <list>
-- route narration: <one line describing the route>
-
-### Scene 1: <title>
-- clips: <filenames with time ranges>
-- narration: <first-person line, 10-40 words>
-- mood: <mood word>
-- duration: <estimated seconds>
-
-### Scene 2: <title>
-...
-
-### Scene N: <title>
-
-Dropped clips:
-- <filename>: <reason>
-
-Title suggestion: <3-15 words>
-```

package/skills/markcut/docs/templates/vlog/prompts/review-fix.md

@@ -1,27 +0,0 @@
-# Prompt: Apply Reviewer Findings
-
-> Fill every `{placeholder}`, then execute in your own context.
-> Used after a FAIL verdict. You edit; the reviewer never does.
-
----
-
-Apply the reviewer findings below to `{vlog_md_path}` with minimal, targeted edits.
-
-Reviewer findings (JSON):
-
-{findings_json}
-
-Rules:
-- Address every `blocker` and `major` finding. Minor findings only when trivially fixable.
-- Each fix must respect TEMPLATE.md §2 scene grammar and §3 authoring rules.
-- Narration fabrication → rewrite the line to use conservative paraphrase of clip captions.
-- Broken src → fix the path or add a placeholder image.
-- Duration mismatch → shorten/lengthen narration and adjust subtitle `duration:`.
-- Blank frames → check startFrom/endAt values, clip ordering, and file existence.
-
-After editing, list what changed:
-```
-- <finding id>: <scene> — <one-line description of the edit>
-```
-
-Then re-render and re-review.

package/skills/markcut/docs/templates/vlog/prompts/storyboard.md

@@ -1,41 +0,0 @@
-# Prompt: Vlog Scene → Markcut Markdown
-
-> Fill every `{placeholder}`, then execute in your own context.
-> Run once per scene in the outline. Constraints mirror TEMPLATE.md §2/§3.
-
----
-
-You are writing one scene of a {style} vlog. The scene will be rendered with markcut markdown syntax.
-
-Theme: {theme}
-Creator: {creator_profile}
-Language: {language}
-
-Scene from outline:
-
-{outline_scene}
-
-All clips are in the media directory `{media_dir}`. Use filenames as `src:` values relative to this directory.
-
-**A. Media selection:**
-- 1-4 clips for this scene. Use imaging for photos, videos for clips. See `startFrom`/`endAt` for trimming.
-- Image: `- image src:{filename} duration:{s} start:{s}`
-- Video: `- video src:{filename} startFrom:{trim_start} endAt:{trim_end} start:{s}`
-- First clip starts at `start:0`. Subsequent clips start at the cumulative previous duration.
-
-**B. Narration:**
-- One `subtitle script:"..."` node per scene. Duration matches total scene length.
-- First-person, conversational, matching creator profile.
-- Every word grounded in clip captions, user labels, or GPS. No fabrication.
-- Length: 10-40 words (~3-15 seconds at 2.5 w/s).
-- No parentheses, no ellipses. At most one emoji.
-
-Output format for ONE scene:
-
-```
-## <SceneTitle (PascalCase)>
-layout:parallel
-- image src:<filename> duration:<s> start:<s>
-- video src:<filename> startFrom:<s> endAt:<s> start:<s>
-- subtitle script:"<narration line>" duration:<total_s>
-```

package/src/render/cli.test.ts

@@ -1,13 +0,0 @@
-import { describe, expect, it } from "vitest";
-import { parseArgs } from "./cli.mjs";
-
-describe("parseArgs", () => {
-  it("supports render without an aspect flag", () => {
-    const args = parseArgs(["node", "markcut", "render", "input.md", "--output", "out/video.mp4"]);
-
-    expect(args.command).toBe("render");
-    expect(args.file).toBe("input.md");
-    expect(args.output).toBe("out/video.mp4");
-    expect(args).not.toHaveProperty("aspect");
-  });
-});