The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes content from a user's Obsidian wiki, which represents a surface for indirect prompt injection if the wiki contains instructions designed to manipulate the agent.
Ingestion points: The skill reads page summaries, frontmatter, and full page bodies (Steps 2-4) from the local vault directory.
Boundary markers: The skill includes a clear 'This skill is READ-ONLY' section and explicit instructions to only propose changes rather than execute them, which serves as a behavioral boundary for the agent.
Capability inventory: The skill utilizes shell-based tools including obsidian-wiki, qmd, and grep. It has the ability to append to a log file (log.md) via the obsidian-wiki memory log command.
Sanitization: No explicit sanitization or escaping of the wiki content is mentioned before it is processed by the LLM.
[COMMAND_EXECUTION]: The skill relies on executing external CLI tools (obsidian-wiki and qmd) to perform graph queries and semantic searches. While these are core to the skill's functionality, they involve spawning subprocesses using paths derived from environment variables like OBSIDIAN_VAULT_PATH and QMD_CLI.

wiki-query