competitor-research
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection (HIGH): The skill ingests untrusted data from arbitrary external websites and possesses file-write capabilities, creating a high-risk surface for indirect injection attacks.
- Ingestion points: The skill uses
WebFetchin Steps 2, 3, and 4 to retrieve data from LinkedIn, G2, Product Hunt, and company pricing pages. - Boundary markers: Absent. There are no delimiters or instructions to the agent to treat the fetched content as data rather than instructions.
- Capability inventory: The skill uses the
Writetool to save reports to the local filesystem at~/.claude/cache/competitor-reports/. - Sanitization: Absent. The fetched data is interpolated directly into a markdown report and written to disk without validation.
- File System Access (MEDIUM): The skill writes files using names derived from potentially untrusted sources.
- Evidence: The file naming convention
{公司名}-{日期}.mdrelies on the company name, which could be manipulated by an attacker to attempt path traversal (e.g.,../../../filename.md) if the name is extracted from a malicious website's metadata.
Recommendations
- AI detected serious security threats
Audit Metadata