The Agent Skills Directory

[PROMPT_INJECTION]: The skill incorporates instructional directives to ensure autonomous operation, such as "Work autonomously" and "Do NOT ask the user questions." These are common behavioral constraints in agent skills intended to reduce conversational overhead.
[COMMAND_EXECUTION]: The skill performs local environment discovery by searching the ~/.claude/projects/ directory to identify project paths for the purpose of logging telemetry data to skill-telemetry.md. This is a localized operation within the platform's configuration space.
[COMMAND_EXECUTION]: The skill provides templates and instructions for the agent to execute system commands (e.g., git log) and to use the npx command for running the registry's CLI tool, which are necessary for its stated development and validation functions.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the @skills-hub-ai/cli package from the npm registry. This package is the official tool for the skill's target registry (skills-hub.ai) and is managed by the skill's authoring organization.
[PROMPT_INJECTION]: The skill defines an indirect prompt injection surface by ingesting user descriptions and using them to generate instruction sets for new skills.
Ingestion points: User-provided skill description in Phase 1 (SKILL.md).
Boundary markers: The generated output uses clear markdown headers and phase separators.
Capability inventory: The agent has capabilities for bash command execution and file system interaction.
Sanitization: No explicit content sanitization is performed on the user description before instruction generation.

skill-writer