The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions utilize shell commands (find, bash, bun) to locate and execute its core logic scripts and manage local dependencies. This is the standard operational method for the skill.
[EXTERNAL_DOWNLOADS]: The skill connects to Amazon Photos infrastructure (a well-known service) to download photo and video content via the Amazon Drive JSON API. It uses Playwright to perform these requests within a browser context to maintain session consistency.
[PROMPT_INJECTION]: The script scripts/download-album.ts is vulnerable to indirect prompt injection or data poisoning via the ingestion of untrusted metadata.
Ingestion points: The script reads the n.name field from the Amazon Drive API response (JSON) which represents the names of files in the shared album.
Boundary markers: Absent. There are no delimiters or warnings for the agent or file system when processing these names.
Capability inventory: The skill uses mkdir and writeFile in scripts/download-album.ts to create the output directory and save the downloaded files.
Sanitization: Absent. The skill does not sanitize the n.name property (e.g., checking for directory traversal sequences like ../). A malicious user could craft a shared album with file names designed to overwrite sensitive files outside the intended output directory.

amazon-photos-album-download