Agent Skills guide

An Agent Skill is a set of task instructions for an AI, sometimes with examples, references, and executable scripts. It helps the AI follow a repeatable workflow. It is not a model, database, or account, and it does not guarantee correct results.

TianGong LCA Skills organize tasks; the tiangong-lca CLI performs platform operations and manages local sign-in. Use the CLI yourself or let a Skills-capable AI client invoke it. Installing a Skill grants no additional rights and does not require MCP setup.

Start with one read-only skill

1. Install a search skill.
2. Run your first flow search and inspect the result.
3. Choose further tasks using the Skill catalogue and prerequisites.

The introduction uses flow-hybrid-search, which only reads data. You will retain request, preview, and actual response files and distinguish no matches, missing sign-in, and execution failure. No Skills, CLI, or Data Foundry checkout is required.

Not every catalogue entry can run after copying a single folder. The three hybrid-search skills have complete standalone runtime packages; others may need the full repository, Data Foundry, existing build artifacts, or additional authorization. Check prerequisites before installing.

Without an AI, use the CLI user guide. For local file packages only, use tidas tools.

Read by task

In this section

5 topics

• Install one search skillInstall Flow Hybrid Search in your project and check its files, runtime, and CLI sign-in.
• First task: find a flowAsk an AI to search for carbon dioxide emissions to air, retain the evidence, and independently check execution.
• Skill catalogue and prerequisitesChoose by task and distinguish standalone search packages, full-repository workflows, and Data Foundry requirements.
• Organize tasks safelyDefine inputs, read scope, write permissions, and deliverables so AI work can be checked and reproduced.
• Troubleshooting and updatesDistinguish discovery, runtime, sign-in, and empty-result issues, and safely update or remove one skill.