exam-forecast
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources and processes it within the agent context.
- Ingestion points: Past exams provided as PDFs, pasted text, or file paths in
SKILL.md(Step 1: Intake). - Boundary markers: The instructions do not define clear delimiters or use explicit 'ignore embedded instructions' warnings for the analyzed content.
- Capability inventory: The skill has file-writing capabilities to local paths (
~/.claude/plugins/config/claude-for-legal/law-student/exam-forecasts/). - Sanitization: There is no evidence of sanitization or validation of the content extracted from the exam files before it is used to generate the final report.
- [DATA_EXFILTRATION]: The skill accesses local configuration files and environment metadata.
- Evidence: It reads from
~/.claude/plugins/config/claude-for-legal/law-student/CLAUDE.mdto load class, professor, and syllabus information. - Context: While these are sensitive configuration paths, there are no network operations or external exfiltration patterns detected; data is processed locally to generate a study forecast file.
Audit Metadata