The Agent Skills Directory

[SAFE]: The skill implements a local indexing service with strong security defaults. It includes a secret redaction engine (SECRET_RE) that identifies and masks sensitive patterns such as API keys, tokens, and passwords during the ingestion of session transcripts. The database and configuration directories are created with restricted permissions (0700 for directories, 0600 for the database file) to prevent unauthorized local access. Command executions in the Python and shell scripts are restricted to internal maintenance tasks, such as triggering background indexing and running evaluation scripts using safe subprocess patterns that avoid shell injection. No network activity or external data exfiltration was detected.

recall