The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by fetching and processing arbitrary content from external URLs.
Ingestion points: Untrusted content enters the agent context via WebFetch for URLs and through raw text provided by the user.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are used when processing the fetched content.
Capability inventory: The skill can read/write to ~/Documents/notes/inbox.org, execute the date system command, and trigger other analysis skills (/ljg-xray).
Sanitization: There is no evidence of content sanitization or validation before the data is formatted into Org-mode and saved locally or passed to subsequent tools.
[COMMAND_EXECUTION]: The skill uses the system command date '+[%Y-%m-%d %a %H:%M]' to generate timestamps for its entries. This is a standard utility but involves the execution of a shell command.
[EXTERNAL_DOWNLOADS]: The skill fetches content from arbitrary remote URLs using the WebFetch tool. While this is the primary purpose of the skill, it involves interaction with untrusted external sources.

ljg-clip