inboundfy-pesquisa-cliente
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data and integrates it into project artifacts, creating a vulnerability where malicious instructions in reviews or websites could influence agent behavior.
- Ingestion points: The skill is designed to ingest data from arbitrary URLs, public communities, product reviews, and transcripts provided by the user (SKILL.md).
- Boundary markers: Instructions require the agent to separate literal speech from observations and interpretations, which provides some structural separation but does not explicitly prevent the execution of embedded instructions (SKILL.md).
- Capability inventory: The skill has the capability to read local context files in the
.inboundfy/directory and write updates tofaq.md,base-editorial.md,pesquisa.md, andpersonas.md(SKILL.md, REFERENCIA.md). - Sanitization: No explicit sanitization, filtering, or escaping mechanisms for the ingested external content are described.
- [EXTERNAL_DOWNLOADS]: The skill performs network requests to fetch research material from non-whitelisted public domains and communities (SKILL.md).
Audit Metadata