The Agent Skills Directory

[COMMAND_EXECUTION]: Security Prompt Bypass: The skill instructs the agent to evade platform-level authentication or consent prompts when modifying sensitive directories such as .claude/. It suggests a two-step process: writing to a non-restricted temporary directory (_/) and then using the mv command to move the file to the target restricted location.
[COMMAND_EXECUTION]: Automated File System Interaction: The instructions require the use of discovery tools like Explore Agent, Glob, and Grep to read and analyze local source code, which involves executing automated file-system operations.
[PROMPT_INJECTION]: Indirect Injection Surface: The skill ingests untrusted user descriptions to generate plans without sanitization or boundary markers.
Ingestion points: User task descriptions (SKILL.md).
Boundary markers: Absent.
Capability inventory: Writing to file system, moving files (mv), and searching files (grep, glob).
Sanitization: Absent.

create-plan