The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize external, untrusted sources which creates a surface for indirect prompt injection.
Ingestion points: Reads documents from the raw/ directory and fetches content from user-provided URLs.
Boundary markers: The skill instructs the agent to read and summarize content into markdown pages, but does not provide specific delimiters or instructions to ignore embedded adversarial commands within the sources.
Capability inventory: The skill has the capability to write to the file system (creating and updating markdown files in the wiki/ directory) and perform network reads (fetching URLs).
Sanitization: No explicit sanitization or filtering mechanisms are described to prevent the agent from obeying instructions hidden within ingested text.
[EXTERNAL_DOWNLOADS]: The skill includes functionality to fetch content from user-specified URLs and save them as snapshots in the raw/ directory.
[COMMAND_EXECUTION]: The skill performs extensive file system operations, including creating a directory structure (wiki/, wiki/sources/, wiki/entities/, etc.) and writing/updating multiple markdown files based on processed data.

wiki-ingest