The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently utilizes shell commands to manage the project environment, including pnpm install for dependency management, chmod +x to make scripts executable, and node -e scripts for programmatic modification of JSON configuration files (e.g., in brief-intake.md and tools.md). It also uses nohup to launch long-running background processes such as development servers.
[INDIRECT_PROMPT_INJECTION]: The 'autonomous-inference mode' described in references/brief-intake.md ingests untrusted data from the user's initial launch prompt or the BRIEF.md file to automatically define application entities, fields, and feature lists. This ingestion point lacks explicit boundary markers or sanitization, and is combined with significant capabilities including file-write access, shell command execution, and database provisioning tools.
[DYNAMIC_EXECUTION]: In references/init.md, the skill grants execution permissions to and runs locally created or copied shell scripts (./.constructive/init.sh) to initialize the project state.

constructive-harness