The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill encourages the agent to read and follow instructions from project-level files such as AGENTS.md, CLAUDE.md, and IDE-specific rules. These files represent a surface for indirect prompt injection because they can contain malicious instructions that influence the agent's behavior. ● Ingestion points: SKILL.md lists several files (AGENTS.md, CLAUDE.md, .cursorrules, etc.) as sources of context and instructions. ● Boundary markers: There are no instructions to use delimiters or ignore embedded instructions when reading these files. ● Capability inventory: The skill involves broad capabilities including reading/writing files and performing git operations. ● Sanitization: The methodology does not include steps to sanitize or validate the content of these external files.

vibe-coding