The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script ../obsidian-notes-karpathy/scripts/detect_lifecycle.py to analyze the target vault before performing operations. Running scripts from relative paths outside the immediate skill folder is a security risk if the environment or the scripts themselves are untrusted.
[PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection as it processes content from an existing vault and external reference files to generate AGENTS.md and CLAUDE.md, which provide instructions for future agent behavior. • Ingestion points: Target vault content and shared reference files in ../obsidian-notes-karpathy/references/. • Boundary markers: No specific delimiters or safety warnings for embedded instructions were identified. • Capability inventory: File system creation and modification, plus shell command execution via the detect_lifecycle.py script and obsidian-cli. • Sanitization: No explicit sanitization of the ingested data or validation of the external reference templates is mentioned.

kb-init