Safety Checks

PassAudited by ClawScan on May 1, 2026.

Overview

The visible artifacts describe a local, instruction-only safety-checking skill whose file/state access is disclosed and aligned with its purpose.

This appears safe to install as an instruction-only local safety checker. Before using it, review the optional companion skills separately and be careful with --clear or --clear-state because they may remove local cache or session data.

Findings (3)

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.

What this means

Using the clear options could remove cached data or local state that the user may have expected to keep.

Why it was flagged

The skill exposes user-invoked options that can delete or clear local cache/session state. This is aligned with the safety and hygiene purpose, but it is still a local mutation capability.

Skill content
| --clear | No | Clear stale cache entries | ... | --clear-state | No | Clear any leaked state |
Recommendation

Use clear flags only when you intend to remove stale cache or leaked state, and confirm the configured paths before running them.

What this means

Installing the recommended companion skills may add capabilities beyond this instruction-only package.

Why it was flagged

The skill recommends installing additional related skills for full enforcement integration. This is disclosed and purpose-aligned, but those additional skills expand the reviewed supply chain.

Skill content
**Dependencies**: `leegitw/constraint-engine` ... `openclaw install leegitw/context-verifier` ... `openclaw install leegitw/failure-memory` ...
Recommendation

Review the companion skills and their permissions before installing them, especially if enabling full constraint enforcement integration.

What this means

Safety reports may reveal local configuration, cache, or session-state information inside the workspace output directory.

Why it was flagged

The skill inspects local cache/session state and writes safety results to a persistent workspace directory. This is central to its purpose and disclosed as local-only, but it involves persistent agent context/state.

Skill content
cache staleness detection, and cross-session state checks ... Results are written to `output/safety/` in your workspace.
Recommendation

Keep output/safety/ within a private workspace and avoid committing or sharing safety reports if they include environment details.