external-dataset-curated-import

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external LCA datasets which could theoretically contain malicious data intended to influence the agent's behavior during the curation workflow.
  • Ingestion points: External source packages are ingested and processed via the tiangong-lca dataset import-lca convert command as described in SKILL.md.
  • Boundary markers: The skill requires schema validation and deterministic QA gates through the CLI to verify data integrity.
  • Capability inventory: The agent performs shell command execution using the tiangong-lca tool and manages files within the .foundry/workspaces/ directory.
  • Sanitization: The instructions explicitly forbid manual parsing within the skill, delegating conversion and validation to structured vendor CLI tools (tiangong-lca, tidas-tools).
  • [COMMAND_EXECUTION]: The skill drives the tiangong-lca CLI to manage datasets and authentication states. These operations are the core functionality of the skill and utilize the vendor's official tooling.
  • [SAFE]: The skill implements strong credential safety by explicitly instructing the agent never to request passwords, tokens, or API keys. It directs the user to perform authentication in a trusted terminal using auth login, which is a security best practice.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 05:39 AM
Security Audit — agent-trust-hub — external-dataset-curated-import