The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the bk CLI to interact with Buildkite services, enabling operations like build triggering (bk build create), job management, and pipeline modification.
[REMOTE_CODE_EXECUTION]: The bk agent run command downloads and executes the Buildkite agent binary from the official Buildkite GitHub releases page to run an ephemeral local agent.
[DYNAMIC_EXECUTION]: Enables the installation of additional agent skills from the Buildkite organization's GitHub repository via the bk skill add command.
[CREDENTIALS_UNSAFE]: Provides commands to access sensitive authentication data, such as bk auth token which prints the session token, and bk secret get which retrieves cluster-scoped secrets.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect instructions through data processing.
Ingestion points: Accesses job logs via bk job log and processes external CI configurations during conversion via bk pipeline convert.
Boundary markers: There are no instructions for using delimiters or boundary markers when reading logs or external configurations.
Capability inventory: The skill can perform high-privilege actions including build creation, secret management, and running the Buildkite agent.
Sanitization: The skill lacks explicit instructions for sanitizing or validating the content of job logs or converted pipeline files before the agent acts on them.

buildkite-cli