The Agent Skills Directory

[COMMAND_EXECUTION]: The workflow involves executing several system commands using the joelclaw, granola, and secrets CLI utilities to manage sessions and fetch data.
[DATA_EXFILTRATION]: The skill accesses highly sensitive personal data, including email threads, meeting transcripts, and API tokens (front_api_token), which are then used to build dossiers stored locally.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface by processing data from untrusted external sources.
Ingestion points: Ingests raw content from email threads via joelclaw email read and meeting transcripts via granola meeting --transcript.
Boundary markers: The agent lacks explicit boundary markers or instructions to treat fetched communication content as data rather than instructions.
Capability inventory: The agent has extensive capabilities including shell command execution, credential leasing, and local file system write access.
Sanitization: No sanitization or filtering logic is provided to validate the content of processed emails or transcripts against malicious instructions.

person-dossier