The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the bm (or basic-memory) command-line tool to perform diagnostic checks (bm doctor) and maintenance tasks (bm reindex). These operations are restricted to local projects as described in the skill instructions.
[INDIRECT_PROMPT_INJECTION]: The skill processes structured JSON output from the bm doctor command, which includes suggested fix commands. If project metadata or file contents can influence the tool's output, the agent might be prompted to execute unintended commands. However, the instructions mandate showing the command to the user and obtaining explicit approval before execution, which provides a significant safeguard.
Ingestion points: Shell output from bm doctor --json (SKILL.md).
Boundary markers: None explicitly defined in the prompt instructions for the JSON parsing.
Capability inventory: The agent is authorized to execute commands provided in the command field of the JSON output (SKILL.md).
Sanitization: The skill relies on user review; instructions state to 'Show each command exactly as doctor printed it and ask before running it.'

memory-doctor