plan-gtm
Fail
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: CRITICALEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The file
references/benchmarks-2026.mdcontains multiple URLs that point to domains flagged as malicious or phishing by security scanners. Specifically, the domains associated withsaasmag.comare blacklisted, and those associated withgetmonetizely.comare flagged for phishing activities. While these are listed as benchmarks, their presence in the skill's reference data poses a risk to both the agent and the user. - [INDIRECT_PROMPT_INJECTION]: The skill implements a workflow that ingests untrusted data from repository source code and founder interview logs, creating a vulnerability to indirect prompt injection.
- Ingestion points: Repository inventory data (Area inventory in Step A of
SKILL.md) and founder interview logs (Step B). - Boundary markers: The instructions provide structure by asking the agent to "Observe" and "Interpret" evidence, but these do not serve as robust security boundaries against adversarial content within the audited files.
- Capability inventory: The agent uses file grepping and reading tools to collect data and writes the findings to a local file named
plan-gtm.md. - Sanitization: No explicit instructions are provided to sanitize, escape, or filter the external content before it is processed or interpolated into the final output.
Recommendations
- AI detected serious security threats
- Contains 4 malicious URL(s) - DO NOT USE
Audit Metadata