brainz-calendar-hardened
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection. 1. Ingestion points: Untrusted calendar data is ingested via the
gcalcli agendacommand described inSKILL.md. 2. Boundary markers: No explicit data delimiters or warnings to ignore embedded instructions are used when processing event content. 3. Capability inventory: The skill can create, search, and delete events usinggcalcliinSKILL.md. 4. Sanitization: No validation or escaping of event content is implemented. Behavioral guardrails inSKILL.mdattempt to mitigate exploitation by forbidding network piping and requiring user confirmation for bulk changes. - [COMMAND_EXECUTION]: The skill executes
gcalclibinary commands to perform calendar operations. This is the core functionality and is subject to stated confirmation workflows for deletions and attendee additions. - [EXTERNAL_DOWNLOADS]: The skill identifies
gcalclias a dependency and provides instructions to install it via pip. This is a standard installation method for a well-known utility.
Audit Metadata