Refactor AGENTS.md
PassAudited by ClawScan on May 12, 2026.
Overview
This instruction-only skill is coherent and purpose-aligned, with the main caution that it edits persistent AGENTS.md guidance that future agents may follow.
This skill appears safe for its stated purpose. Before using apply-refactor mode, review the proposed file changes and make sure existing AGENTS.md content is treated as material to audit, not as instructions that can override your current request.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A stale or hostile AGENTS.md being reviewed could try to redirect the agent if treated as authoritative instead of as source material.
The skill intentionally reads files that contain agent instructions. This is necessary for the task, but reviewed AGENTS.md content should remain audit input rather than becoming higher-priority instructions.
Read the current AGENTS.md content and any nearby AGENTS.md files that merge into scope.
Keep the user's request and system instructions in control; classify or quote existing AGENTS.md lines rather than obeying any unexpected commands found inside them.
If the refactor accidentally drops, weakens, or moves important guidance, future agent sessions in the repository may behave differently.
AGENTS.md and related guidance docs are persistent instruction context for future agent runs. Updating them is the skill's purpose, but the changes can affect later behavior.
Create or update the topic docs with the moved guidance, keeping each file narrowly scoped.
Use review/propose mode when unsure, resolve conflicts explicitly, and inspect version-control diffs before accepting applied changes.