The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from public API endpoints and GitHub repositories which could theoretically contain malicious instructions.
Ingestion points: scripts/lapis_client.py fetches JSON and text data from LAPIS API deployments (e.g., lapis.cov-spectrum.org) and the cov-lineages/pango-designation repository on GitHub.
Boundary markers: Output is formatted as Markdown tables, TSV, or JSON for agent consumption.
Capability inventory: The skill utilizes Read, Write, Edit, and Bash tools, allowing for file operations and shell execution based on processed data.
Sanitization: scripts/lapis_client.py implements a sanitize function that strips control characters (U+0000-U+001F, U+007F-U+009F) and collapses multi-line breaks. This prevents external data from forging CLI output structure or injecting malicious control sequences into the agent's context.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to fetch public scientific data and nomenclature files.
Trusted sources: Fetches SARS-CoV-2 lineage designations and notes from the official pango-designation repository on GitHub.
Public APIs: Connects to established genomic surveillance endpoints including lapis.cov-spectrum.org, lapis.genspectrum.org, and lapis.pathoplexus.org.
Data usage: All downloaded content is parsed as structured data (JSON) or descriptive text; no remote content is executed as code.
[SAFE]: The implementation adheres to several security best practices.
No external dependencies: All scripts use only the Python standard library, as specified in the /// script blocks.
Credential safety: The skill requires no API keys or authentication for its public queries, as stated in the SKILL.md and verified in the code.
Integrity checks: scripts/lapis_client.py performs SHA-256 integrity verification for fetched nomenclature files and enforces database version consistency across API requests.

pathogen-variant-surveillance