The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to interact with the environment through shell commands, specifically the qmd CLI and the node runtime. It directs the agent to execute a local bootstrap script (.scripts/qmd-bootstrap.ts) and provides shell pipelines for resolving index names.
[DYNAMIC_EXECUTION]: The skill contains a hardcoded Node.js snippet intended for execution via node -e. This logic is used at runtime to parse vault-manifest.json and calculate a slug from the current working directory to determine the search index name.
[INDIRECT_PROMPT_INJECTION]: As a search and retrieval tool, this skill is designed to ingest and process arbitrary data from the local vault and configuration files.
Ingestion points: The skill reads from vault-manifest.json and searches across all markdown files within the local vault.
Boundary markers: There are no explicit instructions or delimiters defined to separate the retrieved search content from the agent's system instructions, which could allow instructions embedded in vault notes to influence agent behavior.
Capability inventory: The agent has the capability to execute shell commands (qmd, node, grep) and invoke specific MCP tools (mcp__qmd__query, etc.) based on the search results.
Sanitization: The skill includes a positive security control in the provided Node.js snippet, which uses a strict regular expression (/^[A-Za-z0-9][A-Za-z0-9._-]*$/) to validate the index name before it is interpolated into shell commands, effectively mitigating command injection via the manifest file.

qmd