The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and install external tools from well-known registries, including NPM packages (shardmind, @tobilu/qmd) and GitHub repositories (breferrari/obsidian-mind). these are used to bootstrap the vault environment.
[COMMAND_EXECUTION]: The skill utilizes shell commands for environment setup and operation, including npm install, git clone, shardmind install, and qmd for semantic search. It also executes local scripts via node --experimental-strip-types to manage vault metadata.
[REMOTE_CODE_EXECUTION]: The vault template includes specialized logic in the form of lifecycle hooks (e.g., SessionStart, PostToolUse) and subagents (e.g., slack-archaeologist, review-prep) that execute TypeScript and shell scripts within the agent's environment.
[DATA_EXFILTRATION]: The skill facilitates the ingestion of external data from Slack channels and GitHub pull requests to automate incident reporting and performance reviews. While this is an intended feature, it represents an indirect prompt injection surface where the agent processes untrusted external content.
Ingestion points: External data is ingested via /om-incident-capture (Slack URLs), /om-peer-scan (GitHub PR evidence), and the defuddle tool (web-to-markdown extraction).
Boundary markers: The skill documentation mentions "classification and routing hints" provided by hooks, but does not explicitly define sanitization boundaries for the external data.
Capability inventory: The skill has full access to the bash environment and filesystem to write processed data into vault notes.
Sanitization: The PostToolUse hook performs validation on note frontmatter and wikilinks after writes occur.

obsidian-mind