The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using python to run a local script (llm_wiki.py). This includes commands for root discovery (python "<runtime-script>" root discover) and root resolution (python "<runtime-script>" root resolve).
[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install external code if the runtime script is missing. Evidence: npx skills add huajiexiewenfeng/obsidian-llm-wiki --skill '*' --copy --yes.
[REMOTE_CODE_EXECUTION]: The execution of the Python script involves dynamic path resolution based on the skill's installation directory, which creates a pattern where the agent is instructed to run code from a computed location. Evidence: Resolving <runtime-script> by taking the parent of the SKILL.md directory as <skills-root>.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from an Obsidian Vault, including wiki pages and raw source material, which are untrusted data sources.
Ingestion points: Reading wiki/index.md, ingest/index.md, topics/, projects/, and raw/ files as defined in references/query-workflow.md.
Boundary markers: None explicitly defined for the ingestion of note content to prevent obedience to embedded instructions.
Capability inventory: The skill possesses python execution capabilities, file reading, and file writing (Save-Back functionality).
Sanitization: While the skill mentions not exposing sensitive data in the output (references/synthesis-rules.md), it lacks clear sanitization or filtering for malicious instructions contained within the input notes.

obsidian-wiki-query