The Agent Skills Directory

[COMMAND_EXECUTION]: Executes several command-line tools to process media and manage repository state:
ffprobe for video duration and metadata extraction.
identify (ImageMagick) for image dimension and size verification.
git for checking repository status and identifying file changes.
wrangler (Cloudflare) for uploading processed images to the R2 storage bucket 01mvp-public-assets.
ffmpeg for video compression and format conversion recommendations.
rg (ripgrep) and pnpm for validating output content and repository build integrity.
[EXTERNAL_DOWNLOADS]: Performs network operations to verify asset availability and manage remote storage:
Uploads files to a remote R2 bucket using wrangler with the --remote flag.
Verifies the status of uploaded assets via curl requests to the CDN assets.01mvp.com.
[PROMPT_INJECTION]: The skill ingests untrusted external data in the form of SRT subtitle files, which serves as a potential surface for indirect prompt injection.
Ingestion points: Subtitle files (/path/to/video.srt) used as the primary content source.
Boundary markers: None present; the skill is instructed to use subtitle text as the main source for article generation.
Capability inventory: Significant shell access via wrangler, ffmpeg, git, and pnpm (Corepack).
Sanitization: None described; instructions focus on merging and correcting subtitles into natural paragraphs.

mj-video-publish-pack