curriculum-grade-assist
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTION
Full Analysis
- PROMPT_INJECTION (HIGH): The skill is highly vulnerable to Indirect Prompt Injection (Category 8). It ingests untrusted 'Student Work' (File: SKILL.md, Section: Required Inputs) and uses it to perform 'Analyze Work Against Each Criterion' and 'Determine Performance Level'.
- Ingestion points:
student_responseparameter inskill.pyand--submissionflag inSKILL.md. - Boundary markers: None detected. The workflow indicates direct reading of student content without delimiters or instructions to ignore embedded commands.
- Capability inventory: The skill possesses 'write/execute' equivalents in an educational context—specifically generating official 'Grading Reports' and 'Total Points' which could influence student records or downstream 'Gradebook systems'.
- Sanitization: No evidence of sanitization or filtering of the
student_responsebefore it is processed by the AI for analysis. - COMMAND_EXECUTION (LOW): The
SKILL.mdfile documentation showcases a CLI interface (/curriculum.grade-assist) with file system paths (submissions/*.pdf). While the providedskill.pyis a simulation, the documented usage implies high-privilege file system access and batch processing capabilities.
Recommendations
- AI detected serious security threats
Audit Metadata