The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements a dynamic routing mechanism that resolves and loads 'leaf' skills from relative file system paths (e.g., ../../<name>/SKILL.md). Furthermore, the use-skill playbook provides a trigger for executing skills referenced by external links, which involves runtime loading and execution of skill logic from remote sources.
[REMOTE_CODE_EXECUTION]: The orchestration instructions direct the agent to offer a specific installation command (npx skills@latest add prathamdby/skills) to the user if a required module is missing. This involves the execution of code from a package registry to extend the agent's capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill serves as an entry point for workflows that ingest potentially untrusted data, specifically when searching or cloning external repositories via the research-external playbook.
Ingestion points: The skill ingests user-provided goals and external repository content retrieved via the leaf:box module.
Boundary markers: The playbook templates do not define explicit delimiters or instructions to prevent the agent from following commands embedded in the external content.
Capability inventory: The orchestrator manages access to file system modifications (commit), API interactions (gh, make-pr), and external research tools.
Sanitization: The provided files do not include mechanisms for validating or sanitizing retrieved repository content before it is processed by the agent.

prath-mode