comentarios
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process Instagram comments, which are untrusted external data.
- Ingestion points: Comments provided by the user (list, CSV, or pasted text) in SKILL.md.
- Boundary markers: The skill lacks explicit delimited boundaries for the comments, relying on classification logic.
- Capability inventory: None. The skill only generates text (table/CSV) and does not have file-system write, network, or execution capabilities beyond the LLM's standard output.
- Sanitization: The instructions include basic CSV formatting rules (e.g., escaping quotes), but do not specify measures to prevent instructions embedded within comments from influencing the agent's behavior during classification.
Audit Metadata