The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to coordinate meetings by processing external data such as emails and calendar events, which are known vectors for indirect prompt injection.
Ingestion points: Untrusted data enters the context via email and calendar content (inferred from @openclaw/email and @openclaw/calendar dependencies in package.json).
Boundary markers: Absent; no source code or prompts were provided to verify the use of delimiters or 'ignore' instructions.
Capability inventory: The skill uses sqlite3 for local data persistence and interacts with external services via OAuth and email/calendar providers.
Sanitization: Absent; no implementation logic was provided to verify if external content is escaped or validated.
[No Code Provided] (SAFE): Analysis was limited to package.json and tsconfig.json. No executable scripts or primary agent instructions (SKILL.md) were provided for a full behavioral audit.
[Dependencies] (SAFE): Dependencies consist of local file references and the standard sqlite3 library. No suspicious remote packages or execution patterns were detected.

meetings