The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use Bash to check the ZEPHYR_BAKE_LOG_DIR environment variable at call time to determine the active workspace.
[DATA_EXPOSURE]: The skill targets a specific hardcoded local path (/Users/mtm/Documents/Obsidian Vault/) and operates on files within an Obsidian vault, which typically contains sensitive personal notes and data.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from speech-to-text transcriptions to update bake logs.
Ingestion points: The skill reads "Zephyr" entries extracted from voice memo exports or existing bake logs.
Boundary markers: It uses the sentinel word "Zephyr" and specific date/suffix patterns as markers, but lacks instructions to escape or sanitize the content within these entries before interpolation or processing.
Capability inventory: The agent is authorized to read and write files (.md logs) and check environment variables using Bash.
Sanitization: No security-focused sanitization is present; the "Transcription Errors" logic only addresses linguistic corrections (e.g., "dial" to "dough").

zephyr