skill-improver

Pass

Audited by Gen Agent Trust Hub on Feb 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local shell commands (ls, mv, mkdir, find, grep) to manage feedback files and create summaries within the .claude/ directory.
  • [PROMPT_INJECTION]: Indirect prompt injection risk: The skill's primary function involves reading external feedback data and using it to rewrite other skill files.
  • Ingestion points: Processes files matching .claude/feedback/retro-*.md, which are intended to contain user-provided or system-generated feedback.
  • Boundary markers: There are no specified boundary markers or instructions to disregard malicious directives embedded within the feedback files.
  • Capability inventory: The skill can read/move files and modify the text content of other skills, which directly influences agent behavior.
  • Sanitization: The process lacks a validation or sanitization step to ensure that the feedback does not contain adversarial instructions meant to be 'learned' by other skills.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 25, 2026, 09:36 PM