The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to manage files, calculate SHA256 hashes, and interact with the obsidian-headless CLI for vault synchronization. These are standard operations for a local file management utility.
[EXTERNAL_DOWNLOADS]: The skill references standard package registries (NPM) for installing a helper utility (obsidian-headless). It also references a known open-source repository on GitHub (atomicmemory/llm-wiki-compiler). These are well-known services and do not involve untrusted remote code execution.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external data (articles, papers, web extracts) and synthesize them into markdown pages. While this constitutes an attack surface, the instructions include safety practices such as provenance markers, confidence ratings, and explicitly flagging contradictions to prevent the agent from blindly adopting instructions found in raw data.
[DYNAMIC_EXECUTION]: The linting operation suggests using a Python script via execute_code to scan for orphan pages and validate links. The provided script is a template for local file analysis and does not involve deserializing untrusted data or executing network-sourced payloads.
[DATA_EXFILTRATION]: While the skill interacts with obsidian-headless (which performs vault synchronization), this is a user-configured feature for personal backup/sync between their own devices, not an exfiltration vector to an unknown third party.

llm-wiki