The Agent Skills Directory

[PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by design. In the file beans-prime.ts, the output of the beans prime command is captured and injected directly into the agent's system prompt and context arrays.
Ingestion points: The beans prime shell command output in beans-prime.ts.
Boundary markers: Absent; the content is pushed to output.system and output.context without delimiters or instructions to ignore potential commands within the data.
Capability inventory: The skill possesses the ability to create, list, update, and delete files within the .beans/ directory, and can execute complex GraphQL queries against the task graph.
Sanitization: No sanitization is performed on the CLI output before it is added to the system prompt.
[COMMAND_EXECUTION]: The plugin file beans-prime.ts executes shell commands using a shell-tag literal ($), specifically running which beans, test -f, and beans prime to integrate project context into the AI session.
[EXTERNAL_DOWNLOADS]: The SKILL.md file and its metadata instructions suggest installing the beans binary from a third-party GitHub repository (hmans/beans) using Homebrew or Go.

jb-beans