The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to find and execute scripts within the user's home directory (~/.claude and ~/eon).
Evidence: The skill executes bun and bash on paths generated by find, such as bun "$(find ~/.claude ~/eon -path '*/amazon-photos-album-download/scripts/download-album.ts' | head -1)".
[DYNAMIC_EXECUTION]: The execution of scripts from computed paths found via shell traversal is a risky pattern that could lead to unauthorized code execution if the directory structure is manipulated.
Evidence: Scripts like download-album.ts and make-bundle.sh are loaded from dynamically resolved locations.
[EXTERNAL_DOWNLOADS]: The skill downloads content from Amazon Photos share links.
Evidence: Fetches photo originals from URLs matching https://www.amazon.ca/photos/share/<id>.
[DATA_EXFILTRATION]: The skill reads local photo files (which it notes may contain sensitive EXIF data like GPS, plates, or VINs) and publishes them to external platforms (Cloudflare Workers and GitHub Gists).
Evidence: The pipeline involves downloading originals to $HOME/.cache/album/originals, bundling them, and then using cloudflare-workers-publish and gh gist edit to share the content.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external URLs, creating a surface for indirect prompt injection.
Ingestion points: The ALBUM_URL parameter targets external Amazon Photos albums.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore embedded commands in photo metadata or names.
Capability inventory: The agent has access to Bash, Write, gh CLI, and Cloudflare deployment tools.
Sanitization: There is no evidence of sanitization for the downloaded content before it is processed by the conversion and bundling scripts.

photo-gallery-delivery