The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The ingest_document tool in scripts/graph.py supports downloading files from remote URLs (such as arXiv research papers) using urllib.request.urlretrieve. The downloaded content is stored in the project's local data directory for RAG processing.
[COMMAND_EXECUTION]: The skill executes the ripgrep (rg) search utility using subprocess.Popen in scripts/best_practices.py and scripts/search/keyword.py to efficiently scan markdown and code files within the project root.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes and retrieves content from untrusted external sources (Category 8).
Ingestion points: Untrusted data enters the context via ingest_document (from URLs or local files), update_knowledge_base, and create_knowledge_entry.
Boundary markers: No specialized delimiters or 'ignore internal instructions' markers are applied to retrieved knowledge before it is provided as context to the agent.
Capability inventory: The skill has access to network operations (downloads), file system operations (read/write in project data), and local command execution (ripgrep).
Sanitization: No sanitization, content filtering, or validation is performed on the ingested text or document content.

knowledge