The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data (user-pasted comments and DMs) for analysis, which represents a potential attack surface.
Ingestion points: Found in SKILL.md where user content is interpolated via the [paste] placeholder.
Boundary markers: The skill does not currently use specific delimiters (like XML tags or triple quotes) to separate user data from the instructions, nor does it include explicit 'ignore embedded instructions' warnings.
Capability inventory: The skill documentation does not request or list tools for network access, file system modifications, or shell execution, which significantly restricts the potential impact of an injection.
Sanitization: No automatic sanitization or filtering of the pasted content is implemented within the provided instructions.

comment-mining