The Agent Skills Directory

[SAFE]: The skill uses local shell commands such as find, grep, awk, and stat to analyze markdown files within a user-defined Obsidian vault directory. All operations are restricted to the local file system and are consistent with the skill's stated purpose of housekeeping and metadata analysis.
[DATA_EXFILTRATION]: No unauthorized network operations or data transfers were detected. The skill generates a local digest file in /tmp/hermes-digest.md and outputs the result to the standard output. References to external notification services like Telegram are provided as examples for platform-level cron configuration rather than being implemented as executable commands within the skill's script.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from the vault (markdown files) which may include untrusted data if the vault contains web clippings. This represents an ingestion surface for indirect prompt injection; however, the script's logic is primarily focused on metadata extraction (tags, link existence) rather than interpreting or executing instructions found within the vault content, posing a minimal security risk.

daily-review