ralph

Warn

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a bash script located at .claude/scripts/ralph.sh. Because the script's content is not provided for analysis, its potential actions on the host system cannot be fully audited.
  • [COMMAND_EXECUTION]: The skill depends on and invokes the amp CLI and jq for environment pre-checks and process management.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. Ingestion point: Untrusted data enters via user-provided Markdown PRDs. Boundary markers: Absent. Capability inventory: The agent can write files, commit to Git, and execute bash scripts (SKILL.md). Sanitization: Absent. Malicious instructions in a PRD could be executed as valid tasks.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 14, 2026, 08:43 AM