The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill processes untrusted data from user-provided Obsidian vaults to generate entity classifications and poetic descriptions. Ingestion points: scripts/extract_entities.py walks the vault directory and reads .md file contents while SKILL.md Step 2 reads the resulting scan and samples note content. Boundary markers: No explicit delimiters or instructions are used to separate user data from agent instructions during classification. Capability inventory: The skill executes a local Python script and writes output files including vault-scan.json and temple-data.json. Sanitization: No sanitization of note content is performed before processing by the LLM.
[COMMAND_EXECUTION]: Execution of local utility script. The skill invokes python3 to run scripts/extract_entities.py for vault scanning. The script is bundled with the skill and performs local file read operations limited to markdown files.
[EXTERNAL_DOWNLOADS]: The visualization template fetches Three.js and MediaPipe libraries from well-known services. It uses cdn.jsdelivr.net for Three.js and storage.googleapis.com for MediaPipe models.

temple-generator