event-speaker-kit
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes a local helper script,
scripts/text_check.py, to perform deterministic validation of dates, character counts, and dash usage in generated drafts. This script uses standard Python libraries and performs validation on its inputs to ensure it only processes intended file paths or text strings. The execution is controlled by the agent as part of the internal workflow steps defined inSKILL.md.- [INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface for untrusted data, as it is designed to fetch event pages, read uploaded transcripts, and process user-provided links (Step 2 inSKILL.md). However, it implements high-quality mitigations: - Ingestion Points: Data enters via
web_fetch,web_search, and file-read operations. - Boundary Markers:
SKILL.mdexplicitly instructs the assistant to treat all external content as data, never as instructions, and provides a specific reporting mechanism (a one-line disclaimer at the top of the output) if injection attempts are detected. - Capability Inventory: The skill can execute shell commands for text validation and perform network requests for fetching event details.
- Sanitization: The skill includes a 'scam detection' logic that scans for requests for PII (Personal Identifiable Information), tax IDs, or bank details in the ingested data, flagging them as potential risks to the user.
Audit Metadata