roast-my-resume

Warn

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: In SKILL.md (Section 0.3), the skill instructs the agent to run various shell commands (pdftotext, textutil, pandoc, unzip) on files provided by the user. If the filenames or paths are not sanitized to remove shell metacharacters, it could lead to arbitrary command execution.
  • [DYNAMIC_EXECUTION]: The skill includes a fallback for PDF processing using python3 -c "import pypdf..." (SKILL.md, Section 0.3), which involves dynamic execution of Python code from the shell.
  • [DATA_EXFILTRATION]: SKILL.md (Section 4) directs the agent to verify links extracted from the resume using curl -sI. This functionality can be exploited to conduct Server-Side Request Forgery (SSRF) or to trigger external callbacks when the agent processes a malicious resume containing attacker-controlled URLs.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from various file formats and pasted text, creating an attack surface for indirect prompt injection.
  • Ingestion points: External resume files and user-pasted text are ingested as defined in SKILL.md (Section 0.3).
  • Boundary markers: No explicit delimiters are used to wrap the untrusted resume content, increasing the risk that the agent may interpret text within the resume as its own instructions.
  • Capability inventory: The skill possesses capabilities including shell command execution, network requests via curl, and the ability to write files to the $CRACKED_DIR/resume/ directory (SKILL.md, Section 6).
  • Sanitization: The skill does not define any sanitization or validation steps for the data extracted from the resumes before it is processed.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 10, 2026, 06:30 AM