The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users and agents to install the beans CLI tool from a community-maintained GitHub repository (hmans/beans) using package managers like Homebrew (brew install hmans/beans/beans) or Go (go install github.com/hmans/beans@latest).
[COMMAND_EXECUTION]: The core functionality of the skill relies on executing the beans CLI utility to perform file system operations within the .beans/ directory, including creating, listing, and updating markdown files.
[INDIRECT_PROMPT_INJECTION]: The skill defines a triage workflow that involves reading and interpreting instructions from the bodies of "beans" (tasks or bug reports). Since these can be populated from external sources like user input or automated error reports, they represent an injection surface if the agent treats embedded instructions as authoritative.
Ingestion points: Data enters the agent's context through beans list and beans show commands reading task bodies in SKILL.md and references/cli.md.
Boundary markers: No explicit delimiters or "ignore embedded instructions" warnings are defined for the task body content.
Capability inventory: The skill executes shell commands via the beans CLI and performs file reads/writes in the repository.
Sanitization: No sanitization or validation of the content within the task bodies is specified before processing by the agent.
[DATA_EXPOSURE]: During the triage phase described in SKILL.md, the skill instructs the agent to scan agent-specific configuration paths (.agents/skills/ and ~/.agents/skills/) to discover and match relevant skills to tasks.

beans