The Agent Skills Directory

[COMMAND_EXECUTION]: The skill revolves around the execution of the tokener CLI tool and its various subcommands. The instructions guide the agent on how to use search, auth, and agent harnesses to interact with the Tokener Gateway.
[EXTERNAL_DOWNLOADS]: The commands tokener skill install and tokener update allow for fetching and installing external components or binary updates from console.tokener.dev. These operations are protected by instructions requiring user confirmation or dry-runs.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data retrieved from the Tokener command catalog and search results.
Ingestion points: Command metadata and search results from tokener search and tokener commands show, including fields like notes and known_errors (referenced in references/catalog.md).
Boundary markers: The skill encourages the use of JSON output (--json) to help the agent parse responses reliably, though it lacks explicit delimiters for instructions embedded in data.
Capability inventory: The agent can execute state-changing CLI commands, manage API keys, and install new skills.
Sanitization: The instructions require the agent to inspect the mutation type and perform a --dry-run or seek user confirmation for any operation that is not a simple 'read'.
[CREDENTIALS_UNSAFE]: The skill provides guidelines for handling Personal Access Tokens (PATs) and API keys. It explicitly instructs the agent to treat these as credentials and avoid writing them to logs or shell history, demonstrating a secure handling pattern.
[DYNAMIC_EXECUTION]: The tokener agent command launches external harnesses (e.g., claude, codex), and the execution process fetches a runtime_schema which dynamically defines the command's behavior at runtime.

tokener