The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently executes local scripts and shell commands. Specifically, it invokes python3 ~/ObsidianVault/scripts/update_indexes.py after ingestion and linting operations, and uses ~/ObsidianVault/scripts/llm_wiki_q_timer.sh for performance tracking. It also relies on shell commands like grep, cat, mv, and mkdir to manage and search vault content.
[PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources, including documentation folders in ~/repos/ and web clippings in ~/ObsidianVault/raw/Clippings/. This content is untrusted and presents a surface for indirect prompt injection.
Ingestion points: Markdown files located in ~/ObsidianVault/raw/ and ~/repos/**/docs/**/*.md.
Boundary markers: The instructions do not define clear delimiters or "ignore instructions" directives to isolate the agent from potentially malicious instructions embedded in the source material.
Capability inventory: The agent possesses capabilities to write to the wiki, move files, and execute local scripts, which could be abused if an injection is successful.
Sanitization: No explicit sanitization or filtering logic is specified for the external content before it is processed into summaries or integrated into the knowledge base.
[DATA_EXFILTRATION]: The skill facilitates broad read and write access to the user's entire Obsidian Vault and all git repositories under ~/repos/. While this is part of its core functionality, it creates a significant exposure surface for personal and professional data.

llm-wiki