The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses standard system utilities to perform logic. Specifically, in catalog/capture-daily.md, it executes date +%Y-%m-%d to determine the correct filename for the daily note. It also uses grep in catalog/query.md to search through local note files.
[INDIRECT_PROMPT_INJECTION]: The query.md workflow enables the agent to read and synthesize information from files stored in ~/note/brain/. If an attacker places a malicious markdown file in this directory, the agent could be manipulated during the reflection or synthesis process.
Ingestion points: Files within ~/note/brain/ (raw, daily, wiki, and refs directories) read via local file access or the grep command.
Boundary markers: Absent; there are no specific instructions or delimiters used to ensure the agent treats the content of the notes as untrusted data rather than instructions.
Capability inventory: File system writes to ~/note/brain/, execution of CLI commands (date, grep), and invocation of tools from the brain-hindsight MCP.
Sanitization: While the skill includes a sanitization step for writing guides (redacting API keys in catalog/capture-guide.md), it lacks sanitization or validation logic for content read from the notes during search and synthesis workflows.

brain-lgnh