Planning Validator

ReviewAudited by ClawScan on May 17, 2026.

Overview

This skill appears non-destructive, but it overstates its plan-safety validation compared with the actual included code, which could make users or agents trust unsafe plans.

This skill does not appear to exfiltrate data or perform dangerous actions, but its safety claims exceed the included implementation. Install only if you understand it is a limited validator and continue manually checking tools, permissions, dependencies, and user approval before executing important plans.

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.

What this means

A user or agent may believe this skill has meaningfully validated a plan before execution when the actual validation is very limited.

Why it was flagged

The skill claims several safety mechanisms are implemented, but the provided source only checks for a few required JSON fields and does not implement explicit tool schemas, permission checks, dependency checks, constraint enforcement, watchdogs, or user approval.

Skill content
- [x] Multi-agent validator between steps
- [x] Explicit tool schemas (capabilities + limitations)
- [x] Constraint verification
- [x] Clarification before assumptions
- [x] ICS watchdog reinforced
- [x] CLW watchdog (pattern detection)
- [x] User approval for complex plans
Recommendation

Treat this as a lightweight checklist or prototype only; do not rely on it as a safety gate unless the missing validations are implemented and tested.

What this means

Plans could receive a false sense of feasibility even when they contain unavailable tools, missing permissions, or unrealistic steps.

Why it was flagged

The code presents a realism check but always returns a positive result with high confidence, regardless of the plan contents.

Skill content
def check_reality(self, plan_data):
        """Check if plan is realistic"""
        return {'realistic': True, 'confidence': 0.95}
Recommendation

Replace the hardcoded result with real feasibility checks, and clearly label any placeholder behavior until implemented.