The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external bioinformatics data files (e.g., FASTQ, BAM, FASTA) and structured registry metadata, which could serve as vectors for indirect prompt injection.
Ingestion points: Workflow parameters utilizing LatchFile and LatchDir types in SKILL.md and references/workflow-creation.md allow the agent to interact with data from the latch:/// cloud storage scheme.
Boundary markers: The provided instructions and code examples do not include explicit boundary markers or delimiters to differentiate between data and instructions when processing these external files.
Capability inventory: The skill utilizes the latch CLI for workflow registration (latch register) and execution (latch execute). It also provides SDK functions for file system access and secret retrieval (get_secret).
Sanitization: No specific sanitization or validation logic is detailed for the content of processed files before they are handled by the workflow logic.
[COMMAND_EXECUTION]: The skill documents and encourages the use of the latch command-line interface for managing bioinformatics workflows. Key commands include latch login for authentication, latch init for project setup, latch register for deploying containers, and latch execute for running pipelines.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the latch SDK from PyPI using standard package managers. It also directs users to official documentation, GitHub repositories, and community support channels maintained by LatchBio.

latchbio-integration