google-slides-review
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill accesses Google Slides presentation data and reads local state snapshots from vendor-specific cache directories (
~/.tdx/cache/google-slides-review/). This file access is limited to managing presentation data during the review process. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external Google Slides presentations, including text content and visual thumbnails. This content is interpolated into sub-agent prompts without explicit boundary markers or sanitization. While the sub-agents are constrained to specific reporting tasks, this represents a standard ingestion surface for indirect prompt injection from malicious presentation content.
- [COMMAND_EXECUTION]: The skill generates instructions and recommended tool calls (e.g.,
google_slides_replace_text,google_slides_batch_update) for an orchestrating agent to execute. These commands are specific to the Google Slides domain and are used to remediate the errors found during the review.
Audit Metadata