The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (LOW): The skill documentation instructs the user to install the latch Python package using pip or uv. This is a necessary dependency for interacting with the Latch platform.
[COMMAND_EXECUTION] (LOW): The skill utilizes several CLI commands (latch login, latch init, latch register, latch execute) to manage the workflow lifecycle. These commands are executed locally by the user to authenticate and deploy code to the Latch cloud infrastructure.
[PROMPT_INJECTION] (LOW): (Category 8: Indirect Prompt Injection) The skill is designed to ingest and process untrusted external data in the form of biological files and registry records.
Ingestion points: Data enters through LatchFile, LatchDir, and latch.registry API calls (Table.get, Record.list).
Boundary markers: The documentation does not explicitly detail the use of delimiters or 'ignore' instructions when processing file content in prompts, though standard SDK usage typically handles data as objects.
Capability inventory: The skill can perform network operations (latch register, execute_workflow), manage secrets (get_secret), and modify registry data (Record.update).
Sanitization: The skill relies on Latch SDK's type-safe annotations and the platform's managed execution environment to mitigate risks.
[DATA_EXPOSURE & EXFILTRATION] (SAFE): While the skill handles sensitive genomic data, it promotes secure practices such as using a dedicated secrets manager (get_secret) for API keys rather than hardcoding them.

latchbio-integration