The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes the Hugging Face CLI (hf papers) for searching and reading research papers, which is a well-known service for literature discovery.
Evidence: Found in SKILL.md and references/source-verification.md where instructions guide the agent to use hf papers search, hf papers info, and hf papers read subcommands for terminology research.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest information from external research papers and project documentation, which constitutes an attack surface for indirect prompt injection.
Ingestion points: Research papers retrieved through the Hugging Face CLI and external project documentation from well-known domains such as kubernetes.io and helm.sh as described in references/expression-research.md.
Boundary markers: The skill requires the agent to record all sources in docs/terminology/references.md and link each glossary entry to its corresponding source record as specified in references/records-and-layout.md.
Capability inventory: The skill is authorized to update markdown files within the project's documentation directory (docs/terminology/) and execute the hf command-line tool.
Sanitization: Instructions include explicit mandates to check for accountable authorship, trace definitions to primary work, and apply heightened scrutiny to AI-related projects to ensure the quality and safety of the curated content.

curate-terminology