Add public social and publishing skills

agent-skills/codex/prepare-social-post-launch/SKILL.md

@@ -0,0 +1,128 @@
+---
+name: prepare-social-post-launch
+description: Prepare a tutorial, sponsored-video, or resource launch across social platforms. Use when the user wants to package source files, write and render a PDF guide, upload public downloads, stage channel-specific social posts, configure an optional comment-keyword DM automation, draft follow-up posts, or audit 4:3, 9:16, and 16:9 video exports for cross-platform distribution.
+---
+
+# Prepare Social Post Launch
+
+Turn one long tutorial into a verified resource bundle and a review-ready social launch. Complete every reversible preparation step, but stop before publishing a social post, activating a DM automation, or uploading the full video unless the user explicitly authorizes that final action.
+
+## Canonical video matrix
+
+Use this distribution contract unless the user overrides it:
+
+| Destination | Aspect ratio | Captions |
+|---|---:|---|
+| X | 4:3 | Burned in |
+| LinkedIn | 4:3 | Burned in |
+| Threads | 4:3 | Burned in |
+| Instagram | 9:16 | Burned in |
+| YouTube | 16:9 | No burned-in captions |
+
+Run `scripts/inspect_video_exports.py` on candidate files before uploading. Do not infer the intended destination from a filename alone.
+
+## Required deliverables
+
+Prepare the applicable deliverables in one scoped bundle:
+
+1. The exact HTML demos shown in the video
+2. A ZIP containing those demos and a short README
+3. A polished PDF article based on the transcript
+4. Public Google Drive links for the ZIP and PDF
+5. Requested social posts staged with the matching exports and final copy
+6. An optional ManyChat automation targeting the next Instagram reel and a specific keyword
+7. Follow-up social drafts with the download links, every resource mentioned, and the full-video link
+8. A manifest recording source files, checksums, public URLs, and any unresolved fields
+
+## Workflow
+
+### 1. Audit the source material
+
+- Read the applicable repository instructions.
+- Inspect the video with `ffprobe` and transcribe the complete long-form version.
+- Create a visual timeline or contact sheet so the article and downloads match the video.
+- Identify every demonstrated HTML page from on-screen evidence, filenames, titles, and modification history.
+- Verify each candidate in the Codex in-app browser. Never use Chrome.
+- Package the exact demonstrated versions. Do not silently substitute a later variant.
+
+### 2. Build the HTML download
+
+- Preserve working standalone files as standalone files.
+- Keep dependent assets beside any multi-file page.
+- Add a README with opening instructions and browser requirements.
+- ZIP the demos and record its SHA-256 checksum.
+- Test the extracted bundle before uploading.
+
+### 3. Write and render the article
+
+- Use the user's requested voice skill and `no-ai-slop` when available.
+- Build the article from transcript evidence, not recollection.
+- Lead with the strongest practical change or surprise.
+- Explain the reusable mechanism with concrete prompts and examples.
+- Disclose sponsorship plainly.
+- Treat price and model claims as time-sensitive. Attribute observed costs to the tutorial and advise readers to check the live estimate.
+- Use the PDF skill to render the final document under `output/pdf/`.
+- Render every PDF page to images and inspect the full contact sheet for clipping, overflow, broken links, and weak hierarchy.
+
+### 4. Upload and verify downloads
+
+- Prefer the Google Drive connector.
+- If its write scope is unavailable, use the signed-in Codex browser.
+- Create a clearly named folder and upload the ZIP and PDF.
+- Set each delivery file to `Anyone with the link` and `Viewer` when the user has asked for public delivery.
+- Read the file IDs or copied links back from the live Drive state.
+- Verify that each item visibly reports shared/public access. Never invent or guess a Drive URL.
+
+### 5. Stage requested social posts
+
+- Match each requested destination to the verified export in the canonical video matrix.
+- Check every crop and media preview visually before advancing.
+- Write short platform-appropriate copy with one mechanism and one useful takeaway.
+- Add a keyword CTA only when a comment-to-DM workflow is part of the request.
+- Include sponsorship disclosure when applicable.
+- For Instagram, turn off Facebook cross-posting unless the user asked for it.
+- Stop at each platform's final share or publish control. Do not publish.
+
+### 6. Configure ManyChat
+
+Use the `Auto-DM links from comments` quick automation when an Instagram comment-to-DM workflow is requested.
+
+- Target `next post or reel`.
+- Match a specific memorable keyword, normally uppercase.
+- Enable public comment replies and write three short, natural variants.
+- Add an opening DM and a clear confirmation button.
+- Deliver direct links to the HTML ZIP, PDF, and full YouTube video.
+- Do not add an email gate unless the user asks for lead capture.
+- Do not enable follow-gating by default.
+- Stop before `Go live`.
+
+If the YouTube URL is missing, leave an explicit `[ADD YOUTUBE URL BEFORE GOING LIVE]` marker and keep the automation inactive. Do not replace it with a different video.
+
+### 7. Draft the follow-up social post
+
+Prepare a separate resource post for the requested follow-up channel after the main video post. It should:
+
+- Open with what is now available.
+- Give one useful lesson from the video.
+- Link the HTML archive, PDF guide, and full YouTube video.
+- Include direct links to the important tools and source references.
+- Stay concise enough to scan on mobile.
+- Remain a draft unless the user explicitly asks to post it.
+
+Use `references/copy-patterns.md` for reusable copy structure.
+
+### 8. Final verification and handoff
+
+- Confirm the local files exist and checksums match the upload sources.
+- Read every social caption back from its live composer.
+- Read the ManyChat trigger, replies, opening DM, and delivery message back from the live builder.
+- Confirm that neither Instagram nor ManyChat is live.
+- Keep relevant social and ManyChat tabs open as handoffs when user action is still required.
+- Report unresolved fields prominently, especially the YouTube URL.
+
+## Safety gates
+
+- Uploading the requested files to the named Drive and staging requested social posts are authorized when the user asked for those actions.
+- Publishing, activating an automation, sending a broadcast, and uploading a YouTube video are separate external actions. Do not infer those permissions.
+- Do not let a placeholder URL reach a live automation.
+- Never expose unrelated files, Drive folders, messages, contacts, or account data.

agent-skills/codex/prepare-social-post-launch/agents/openai.yaml

@@ -0,0 +1,4 @@
+interface:
+  display_name: "Prepare Social Post Launch"
+  short_description: "Prepare resource launches across social platforms."
+  default_prompt: "Use $prepare-social-post-launch to package the resources, validate channel-specific media, stage the requested social posts, and stop before publishing or activating automations."

agent-skills/codex/prepare-social-post-launch/references/copy-patterns.md

@@ -0,0 +1,65 @@
+# Copy patterns
+
+## Social caption with a comment-to-DM CTA
+
+```text
+I used [agent/tool] to build [specific result] from [strong references].
+
+The trick is [mechanism]. [Actionable sequence].
+
+Comment KEYWORD and I'll send you [specific bundle].
+
+Sponsored by [sponsor].
+```
+
+## Public replies
+
+Use three variations so the automation does not feel mechanical:
+
+```text
+Sent! Check your DMs 👀
+Just sent the files. Have fun with them.
+They’re in your DMs — [specific contents].
+```
+
+## Opening DM
+
+```text
+I packaged [exact contents], plus [guide or bonus]. Tap below and I'll send everything.
+```
+
+Button: `Send me the files`
+
+## Delivery DM
+
+```text
+Here you go:
+
+HTML demos: [public link]
+PDF guide: [public link]
+Full video: [YouTube link]
+
+[One immediately useful tip from the tutorial.]
+
+Sponsored by [sponsor].
+```
+
+## Follow-up X post
+
+```text
+The [number] [resources] from my [topic] video are ready.
+
+I also wrote a practical guide to [mechanism and payoff].
+
+HTML downloads:
+[link]
+
+PDF guide:
+[link]
+
+Full video:
+[link]
+
+Resources:
+[short direct list]
+```

agent-skills/codex/prepare-social-post-launch/scripts/inspect_video_exports.py

@@ -0,0 +1,80 @@
+#!/usr/bin/env python3
+"""Inspect video dimensions and map them to the social distribution matrix."""
+
+from __future__ import annotations
+
+import json
+import shutil
+import subprocess
+import sys
+from pathlib import Path
+
+
+def probe(path: Path) -> dict:
+    command = [
+        "ffprobe",
+        "-v",
+        "error",
+        "-select_streams",
+        "v:0",
+        "-show_entries",
+        "stream=width,height,avg_frame_rate:format=duration",
+        "-of",
+        "json",
+        str(path),
+    ]
+    result = subprocess.run(command, check=True, capture_output=True, text=True)
+    payload = json.loads(result.stdout)
+    stream = payload["streams"][0]
+    width = int(stream["width"])
+    height = int(stream["height"])
+    ratio = width / height
+
+    if abs(ratio - 4 / 3) < 0.03:
+        destination = "X, LinkedIn, Threads"
+        expected_captions = "burned in"
+    elif abs(ratio - 9 / 16) < 0.03:
+        destination = "Instagram"
+        expected_captions = "burned in"
+    elif abs(ratio - 16 / 9) < 0.03:
+        destination = "YouTube"
+        expected_captions = "not burned in"
+    else:
+        destination = "unclassified"
+        expected_captions = "review manually"
+
+    return {
+        "file": str(path.resolve()),
+        "width": width,
+        "height": height,
+        "aspect_ratio": round(ratio, 4),
+        "duration_seconds": round(float(payload["format"]["duration"]), 2),
+        "frame_rate": stream.get("avg_frame_rate"),
+        "destination": destination,
+        "expected_captions": expected_captions,
+        "caption_visual_check_required": True,
+    }
+
+
+def main() -> int:
+    if not shutil.which("ffprobe"):
+        print("ffprobe is required", file=sys.stderr)
+        return 2
+    if len(sys.argv) < 2:
+        print(f"Usage: {Path(sys.argv[0]).name} VIDEO [VIDEO ...]", file=sys.stderr)
+        return 2
+
+    records = []
+    for raw in sys.argv[1:]:
+        path = Path(raw).expanduser()
+        if not path.is_file():
+            print(f"Missing file: {path}", file=sys.stderr)
+            return 2
+        records.append(probe(path))
+
+    print(json.dumps(records, indent=2))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())