ai-used-resume
Warn
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on running shell commands through the vibe-resume tool and optional external binaries like claude (for LLM enrichment) and pandoc (for PDF rendering). It also provides scripts for backing up sensitive Claude project data (scripts/backup_claude_projects.sh).
- [EXTERNAL_DOWNLOADS]: The documentation instructs users to clone and install the tool from https://github.com/easyvibecoding/vibe-resume, which is the author's repository. This involves executing third-party code locally.
- [DATA_EXFILTRATION]: The skill accesses highly sensitive files, specifically targeting ~/.claude/projects to extract AI-coding history. It also performs a broad scan of the user's $HOME directory for .git repositories when scan.mode is set to full. This represents a high degree of sensitive data exposure.
- [PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection (Category 8). It ingests untrusted data from AI logs and git commit history which are then used as input for LLM-based enrichment via claude -p. Ingestion points: AI tool logs (Claude Code, Cursor, etc.), git commit messages, and external cloud exports. Boundary markers: Not explicitly defined in the provided instructions, though mention is made of an 'anti-leak rule' in the enrichment prompt. Capability inventory: Command execution via uv run, local file system reads across $HOME, and output to Markdown/DOCX/PDF files. Sanitization: Not detailed in the provided documentation.
Audit Metadata