The Agent Skills Directory

[COMMAND_EXECUTION]: The skill scripts (update-process-references.mjs and verify-process-rows.mjs) act as wrappers for the tiangong-lca CLI tool, delegating tasks such as reference refreshing and row verification to the native executable. It supports specifying a local CLI directory via the --cli-dir argument or TIANGONG_LCA_CLI_DIR environment variable for development purposes.
[INDIRECT_PROMPT_INJECTION]: The skill processes remote process data and local row reports, which represents a potential attack surface for indirect prompt injection if the ingested data contains malicious instructions intended for the agent.
Ingestion points: External process data is ingested via the --rows-file argument in the verification script.
Boundary markers: The skill instructions mandate a 'complete-dataset gate' including strict ProcessSchema.safeParse validation and the preservation of verification artifacts.
Capability inventory: The skill is capable of executing CLI commands through a local launcher and writing reports to directories specified by the --out-dir argument.
Sanitization: The skill delegates input sanitization and logic validation to the underlying tiangong-lca CLI tool and strict schema parsing.
[CREDENTIALS_UNSAFE]: The skill follows security best practices by explicitly prohibiting the agent from asking for, printing, or hardcoding passwords, tokens, or secret keys. It relies on the environment's authenticated OAuth session and directs users to perform sensitive authentication tasks in a trusted terminal.

tiangong-lca-remote-ops