extract

Warn

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes a workflow step that executes a complex shell command using grep and command substitution to search through local directories.
  • [DATA_EXFILTRATION]: The skill specifically targets and reads internal application data stored in $HOME/.claude/projects/, exposing potentially sensitive development history, project logs, and private context to the agent.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by ingesting raw project memory and user descriptions to generate new instruction sets. Evidence Chain:
  • Ingestion points: Project memory files and user-provided pattern descriptions (SKILL.md).
  • Boundary markers: Absent; there are no delimiters or instructions to ignore embedded commands in the source material.
  • Capability inventory: Includes shell command execution (grep) and file system write access for skill creation.
  • Sanitization: The skill verifies formatting but lacks sanitization against malicious instructions embedded in the processed memory data.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 10, 2026, 05:35 AM