paper-plan
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes multiple external and potentially untrusted files such as
NARRATIVE_REPORT.md,STORY.md, and various experiment logs to generate structured paper outlines. It lacks defined boundary markers or sanitization logic to prevent malicious instructions embedded in these files from influencing the agent's behavior during the outline generation process. * Ingestion points:NARRATIVE_REPORT.md,STORY.md,AUTO_REVIEW.md,IDEA_REPORT.md,figures/,findings.md,EXPERIMENT_LOG.md,CLAIMS_FROM_RESULTS.md, and$ARGUMENTS. * Boundary markers: Absent. * Capability inventory:Bash,Write,Edit,Grep,WebSearch, and various MCP tools. * Sanitization: Not specified. - [COMMAND_EXECUTION]: The skill includes a pre-defined bash script that performs git operations, file reads (
cat,awk), and executes local Python tools. Additionally, it explicitly instructs the agent to use shell redirection (cat << 'EOF') to bypass file size limitations when writing to the filesystem without seeking user confirmation. - [DYNAMIC_EXECUTION]: The skill implements a dynamic resolution mechanism to locate and execute a Python script (
extract_paper_style.py). It searches through relative project paths and a path stored in the user's home directory ($HOME/.aris/repo), then executes the found script usingpython3.
Audit Metadata