The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation includes shell commands for document rendering and system configuration.
Evidence in references/schema.md provides commands for pandoc, brew, and tlmgr to convert markdown to PDF or HTML.
The documentation includes a sudo command for updating the TeX Live manager (sudo tlmgr update --self).
Shell commands for searching the vault involve using grep across local directories.
[INDIRECT_PROMPT_INJECTION]: The skill has a high surface area for indirect prompt injection due to its core workflow.
Ingestion points: The capture mode (SKILL.md) fetches content from external URLs, tweets, and articles, saving them to a raw/ folder.
Capability inventory: The skill performs file read/write operations and executes shell commands based on the processed data.
Sanitization: There are no explicit instructions or delimiters used to prevent the agent from following malicious instructions contained within the captured external text during the compile or query phases.
Boundary markers: While the skill separates raw/ and wiki/ content, it does not instruct the agent to ignore instructions embedded in the source material.
[DATA_EXPOSURE]: The skill is designed to interact with the user's local filesystem.
It accesses the directory defined by the environment variable $SECOND_BRAIN_VAULT or defaults to $HOME/Documents/SecondBrain.
It reads configuration files like CLAUDE.md from the vault root and references files in ~/.local/share/makerskills/.

second-brain