The Agent Skills Directory

[COMMAND_EXECUTION]: The skill explicitly instructs the agent to use Bash commands to check the ZEPHYR_BAKE_LOG_DIR environment variable at runtime. While intended for configuration, this pattern encourages the use of shell execution for tasks that might have safer alternatives.
Mitigation: Utilize native platform methods for accessing environment variables rather than executing shell commands.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted text (such as speech-to-text transcriptions of voice memos) from various documents without sanitization.
Ingestion points: The routing logic in ROUTING.md reads from documents containing mixed content to identify and extract entries.
Boundary markers: Absent. The skill identifies entries using the "For Zephyr N" pattern, which does not provide a secure boundary against instructions embedded in the processed data.
Capability inventory: The skill has the capability to read, write, create, and delete file content (via its default move mode) within the local file system, specifically targeting an Obsidian Vault.
Sanitization: No sanitization or instruction-filtering is performed on the body of the extracted entries before they are written to target logs.
Mitigation: Implement strict delimiters for processed data and provide the agent with explicit instructions to ignore any potential commands or directives found within the entry content.

zephyr