skill-first

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • COMMAND_EXECUTION (SAFE): The skill provides a bash command to list installed skills.
  • Evidence: find ~/.claude -path "*/skills/*/SKILL.md" 2>/dev/null | xargs -I{} grep "^name:" {}.
  • Context: This command is used purely for discovery within the agent's specific directory. It only extracts the 'name' field and does not execute the files or pipe content to an interpreter.
  • PROMPT_INJECTION (LOW): The skill uses strong directives to override default task-handling behavior.
  • Evidence: 'Follow exactly
  • Execute the skill's guidance without deviation', 'skill-first is mandatory'.
  • Context: This language enforces the intended workflow of the skill (protocol discipline) and does not target safety filters, system prompt extraction, or adversarial jailbreaking.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:31 PM