The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill accesses documentation and templates located at ../../docs/skill-writing.md and ../../docs/templates/SKILL.md. It also writes newly created skill definitions to skills/<name>/SKILL.md. These file operations are local to the repository and consistent with the skill's stated purpose as a development tool.
[INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by ingesting user-provided descriptions of intended behavior to generate new skill files. If a user provides malicious instructions during the 'Drafting' phase, the resulting skill file could contain prompt injections. However, the skill includes a 'Checkpoint' requiring user confirmation of the draft before the file is saved to the disk.
[COMMAND_EXECUTION]: The skill instructs the agent to guide the user in modifying configuration files (.claude-plugin/plugin.json and .claude-plugin/marketplace.json) to register new skills. This modification is part of the standard deployment workflow for this platform.

write-skill