The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell commands for repository management using Git, automated testing via npx cucumber-js, and project shipping operations. It retrieves specific commands for outbound actions from a local configuration file and performs 'Perturbation' by injecting specific error-throwing statements into production code for verification purposes.
[EXTERNAL_DOWNLOADS]: During project initialization and toolchain setup, the skill is instructed to install packages from official registries. It references well-known development tools including Cucumber, Biome, Playwright, and others.
[DATA_EXFILTRATION]: The skill includes workflows for publishing and releasing code artifacts. These actions are protected by a requirement for explicit human approval before the agent performs any outbound tasks.
[PROMPT_INJECTION]: The skill processes project-specific files such as AGENTS.md and .feature specs which represent an indirect prompt injection surface.
Ingestion points: Reads project-level configuration and documentation files from the repository.
Boundary markers: Uses specific domain-specific language and structured templates to maintain role boundaries.
Capability inventory: Includes file system writes, git operations, package management, and script execution.
Sanitization: Relies on specific workflow logic and user confirmation checkpoints for sensitive outbound tasks.

captain