The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interacts with the system using a dedicated CLI tool (br) to manage task lifecycle and project metadata. It executes commands to list dependencies, create new task beads for fixes or follow-ups, and close project epics.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted data from project artifacts and source code changes (Ingestion points: .beads/artifacts/ and modified source files). While it uses Markdown headers for structure (Boundary markers), it lacks explicit sanitization (Sanitization: Absent) for the data it processes. The skill possesses the capability to modify project state via the br CLI (Capability inventory: br create, update, close), but the risk is mitigated by a mandatory human UAT process and structured sub-agent prompts.

beo-reviewing