The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external LCA datasets (such as BAFU or USLCI), which serves as an ingestion point for untrusted data that could contain malicious instructions.
Ingestion points: External source packages, manifests, and converted dataset rows referenced in SKILL.md.
Boundary markers: The instructions explicitly direct the agent to distinguish between user instructions, AI assumptions, and package source facts.
Capability inventory: The skill utilizes the tiangong-lca CLI for file conversion, queue management, and authentication checks, and it performs file system operations within designated workspace directories.
Sanitization: The workflow mandates the use of deterministic CLI commands, schema validation, and QA reports before data is committed or processed further.
[COMMAND_EXECUTION]: The skill relies on the execution of various subcommands from the tiangong-lca CLI to perform data transformation and validation.
Evidence: Commands such as tiangong-lca dataset import-lca convert, tiangong-lca dataset curation-queue build, and tiangong-lca auth doctor-auth are integrated into the workflow steps.
[SAFE]: The skill demonstrates excellent security hygiene regarding credential management.
Evidence: It contains explicit prohibitions against requesting sensitive information like passwords, API keys, or tokens. It instructs the agent to direct users to perform authentication through the official CLI login process in a separate, trusted terminal environment.

external-dataset-curated-import