koriigami-topic-research
Warn
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: MEDIUMPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [Indirect Prompt Injection] (MEDIUM): The skill processes untrusted external data during its research phase and incorporates it into a local markdown file.
- Ingestion points: Web search results, industry reports, and 'trending conversations' (Step 2).
- Boundary markers: Absent. The skill does not instruct the agent to use delimiters or ignore embedded instructions within the researched content.
- Capability inventory: File system write access (Step 4) and user interaction via the
AskUserQuestiontool. - Sanitization: Absent. Data from external sources is interpolated directly into the output rationales and content gap analysis.
- Multi-step chain risk: The output is designed to be used by the
/koriigami-write-articleskill. A malicious actor could host content that, when researched, injects instructions into the generated topic list to compromise the behavior of the subsequent writing skill. - [Data Exposure & Exfiltration] (LOW): The skill writes output directly to the local filesystem. While this is its primary function, the persistence of unsanitized data from external sources onto a local machine is a minor security concern (Category 2).
Audit Metadata