The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to run bun install -g https://github.com/tobi/qmd, which downloads a package from a third-party repository not associated with the vendor or known trusted entities.
[REMOTE_CODE_EXECUTION]: The global installation of a package directly from a GitHub URL executes the package's installation scripts and provides the resulting binary with execution privileges on the system, which can be used to run arbitrary code.
[COMMAND_EXECUTION]: The skill relies on executing various qmd shell commands to search, retrieve, and update local file content, which grants the agent significant access to the local filesystem.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from local markdown files during search and retrieval operations, which could contain malicious instructions designed to manipulate the agent's behavior.
Ingestion points: Markdown files indexed and retrieved using qmd search and qmd get commands in SKILL.md.
Boundary markers: Absent; there are no instructions or delimiters provided to the agent to distinguish between its instructions and the retrieved document content.
Capability inventory: Execution of qmd CLI commands via the shell and reading of local file contents.
Sanitization: Absent; the tool retrieves and returns full document content to the agent without filtering or sanitization.

qmd