The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest data from untrusted external sources (URLs, YouTube transcripts) and local files (PDFs, images). This creates a vulnerability surface where an attacker can place hidden instructions in a source document to take control of the agent's logic once the content is processed.
Ingestion points: SKILL.md (User-provided URLs, YouTube links, and local file paths passed as arguments to the summarize command).
Boundary markers: Absent; the instructions do not specify any delimiters or warnings to the agent to ignore instructions found within the input data.
Capability inventory: The summarize tool can perform network reads to fetch URL content and local file reads (e.g., /path/to/document.pdf).
Sanitization: Absent; there is no mention of escaping or validating the content retrieved from external sources before the agent processes it.
[COMMAND_EXECUTION]: The skill facilitates the execution of a specific CLI tool (summarize) and uses shell piping with rg (ripgrep) to discover tool capabilities. While these are utility commands, they allow the agent to interact with the underlying host shell to perform file and network operations.

summarize