The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes content from external documentation files without protective boundary instructions or sanitization.\n- Ingestion points: The SKILL.md file instructs the agent to read documentation from paths outside the skill's directory using relative traversal (e.g., ${CLAUDE_SKILL_DIR}/../../../../docs/plugins.md).\n- Boundary markers: Not present. There are no instructions to the agent to treat documentation content as data rather than instructions, or to ignore embedded prompts.\n- Capability inventory: The skill leverages the host agent's file system access; the package contains no executable scripts, binary files, or network-capable tools.\n- Sanitization: None. Content retrieved from documentation files is directly integrated into the agent's reasoning and synthesis process without filtering.

super-help