compose-expert
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/find-composables.shusesgrepto identify@Composablefunctions in the codebase. It uses quoted variables to handle paths and restricts searches to Kotlin files, minimizing command injection risks. - [INDIRECT_PROMPT_INJECTION]: The skill documents the
RichTextParsersystem inreferences/rich-text-parsing.mdfor processing raw content from external Nostr events. - Ingestion points: Raw event content is ingested and parsed by
RichTextParser.kt. - Boundary markers: The system uses a structured segment model (
RichTextParserSegments.kt) to delimit different content types. - Capability inventory: The skill facilitates UI rendering of parsed content and includes a local search utility.
- Sanitization: The
UrlParser.ktutility provides validation for external URLs before they are processed.
Audit Metadata