The Agent Skills Directory

[COMMAND_EXECUTION]: Executes local shell scripts nano-doctor.sh, skill-preamble.sh, and skill-finalize.sh located within the ~/.claude/skills/nanostack/bin/ directory.
[REMOTE_CODE_EXECUTION]: Recommends the execution of npx create-nanostack, which downloads and executes remote code from the npm registry at runtime.
[COMMAND_EXECUTION]: The --fix functionality modifies the agent's .claude/settings.json to insert PreToolUse hooks for Bash and file-editing tools. These hooks function as persistence mechanisms that automatically trigger scripts whenever those tools are invoked by the agent.
[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection via the following characteristics:
Ingestion points: The agent is instructed to process the JSON output from nano-doctor.sh and follow the fix_command provided within that output.
Boundary markers: The instructions lack boundary markers or guidelines to ignore instructions embedded within the script's output.
Capability inventory: The skill possesses capabilities to modify configuration files, change file permissions, and execute shell commands.
Sanitization: There is no evidence of sanitization or validation for the commands and parameters parsed from the diagnostic output before they are suggested to or executed by the agent.

nano-doctor