The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently executes the beaker CLI to manage integrations, verify environments, and trigger runs. This includes commands like beaker run trigger, beaker run status, and beaker agent env check.
[INDIRECT_PROMPT_INJECTION]: The agent processes and relays information from external JSON outputs produced by the beaker CLI.
Ingestion points: Instructions direct the agent to read and act upon data from beaker onboarding status, beaker agent list, and evaluation artifacts like candidates.json.
Boundary markers: The agent is instructed to relay discovered actions "verbatim" to the user, with no specific instructions to treat the data as untrusted or separate from the prompt instructions.
Capability inventory: The skill possesses the capability to execute shell commands and perform Git repository updates (commit/push).
Sanitization: The instructions do not define any sanitization or validation protocols for the external data retrieved from the CLI.
[PRIVILEGE_ESCALATION]: The skill allows the agent to autonomously commit and push integration code to a dedicated repository branch without explicit user confirmation for each operation, representing an elevated level of autonomy.
[EXTERNAL_DOWNLOADS]: The documentation references updating the skill and the beaker-sdk dependency using npx, which downloads and executes packages from the npm registry.

beaker-usage