self-improvement

Pass

Audited by Gen Agent Trust Hub on Apr 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes several local shell scripts (activator.sh, error-detector.sh, extract-skill.sh) to automate its workflows. These scripts perform routine tasks such as checking for error patterns in command outputs, providing reminders to the agent, and scaffolding new skill files based on templates. The logic within these scripts is transparent and restricted to the local workspace.
  • [EXTERNAL_DOWNLOADS]: The documentation references installation methods involving git clone from the author's GitHub repositories and the use of clawdhub. These are standard procedures for managing and installing agent skills and do not involve unauthorized remote code execution.
  • [PROMPT_INJECTION]: The skill is designed to influence agent behavior in future sessions by 'promoting' captured learnings into project-level instruction files like CLAUDE.md and AGENTS.md. While this modifies the agent's operating context, it is the primary intended function of the skill and is carried out through explicit instructions and documented workflows, presenting a low risk of indirect prompt injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 16, 2026, 12:45 PM