The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a standalone Node.js script, okf-validate.mjs, which is used to check bundle conformance. The agent is instructed to execute this utility locally via the command node okf-validate.mjs <bundle>. The script is self-contained and has zero external dependencies.
[EXTERNAL_DOWNLOADS]: The export and enrich commands are designed to fetch external content from URLs, BigQuery datasets, and OpenAPI specifications. This capability is used to extract knowledge substance and transform it into the OKF bundle format, acting as a producer role for the agent.
[PROMPT_INJECTION]: By processing data from external sources like websites and wikis, the skill introduces a vulnerability surface for indirect prompt injection. The skill provides remediation guidance through its 'Adversarial examples' and instructions to re-express substance in the agent's own structure rather than copying raw prose, which helps prevent the accidental execution of embedded instructions.
[DATA_EXPOSURE]: The validation tool okf-validate.mjs reads the local filesystem using standard Node.js APIs to perform its analysis. The skill manifest in permissions.yaml transparently declares this execution capability and specifies that the validator is read-only.

okf