write-research
Fail
Audited by Gen Agent Trust Hub on Jul 12, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Rule 5 in SKILL.md instructs the agent to verify product behavior by running commands like curl or executing sandbox scripts. This behavior mandates high-risk command execution based on instructions found in untrusted external materials.
- [REMOTE_CODE_EXECUTION]: By requiring the agent to execute code to verify claims, the skill introduces a functional path for remote code execution. Malicious research sources can include harmful payloads that the agent is directed to execute as part of its verification process.
- [DATA_EXFILTRATION]: The requirement to use network-enabled tools for verification provides a mechanism for data exfiltration if the agent is directed to a malicious endpoint while its environment context is active.
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. Evidence Chain: 1. Ingestion points: RFCs, library docs, source code, and blogs (SKILL.md Rule 3). 2. Boundary markers: Absent. 3. Capability inventory: Shell access for command and script execution (SKILL.md Rule 5). 4. Sanitization: Absent.
Recommendations
- AI detected serious security threats
Audit Metadata