The Agent Skills Directory

[COMMAND_EXECUTION]: The record.sh script executes several shell commands, including git, qmd, mkdir, and echo. These are used to detect project context via git remote URLs, manage the knowledge base directory structure in ~/.ai-knowledges/, and interface with the qmd CLI for indexing and searching.
[EXTERNAL_DOWNLOADS]: The skill documentation instructs the user to install the @tobilu/qmd package globally using bun or npm. This external dependency is the core component that provides the MCP server functionality.
[PROMPT_INJECTION]: As a knowledge management tool, the skill is subject to indirect prompt injection. Maliciously crafted notes or learnings could be recorded and later retrieved by the AI, potentially influencing its behavior during future interactions.
Ingestion points: Untrusted data enters the system via the record.sh script, which saves user-provided text to markdown files in the project's knowledge directory.
Boundary markers: No specific delimiters or "ignore instructions" markers are defined in the prompt instructions for retrieval.
Capability inventory: The skill utilizes shell commands (git, qmd) and file system write operations.
Sanitization: The skill implements a slugify function to sanitize filenames and a validate_safe_path function to prevent path traversal, ensuring recorded data stays within the intended directory structure.

qmd-knowledge