compliance-policy-check

name	compliance-policy-check
description	Validate planned changes against local framework rules and policy guardrails before implementation or creation.
version	1.1.0
model	sonnet
invoked_by	both
user_invocable	true
tools	["Read","Glob","Grep","Skill"]
agents	["planner","technical-program-manager","reflection-agent","evolution-orchestrator"]
error_handling	graceful
streaming	supported
verified	true
lastVerifiedAt	"2026-02-22T00:00:00.000Z"
source	builtin
trust_score	100
provenance_sha	0b1bc6cde6c722e2

Compliance Policy Check

Overview

Evaluate a design/plan against framework policy and rule constraints before execution. Use this for regulated or high-risk changes.

When to Use

• Before creator workflows for new artifacts
• Before HIGH/EPIC implementation phases
• During reflection when repeated policy violations are observed

Iron Laws

1. NEVER execute or modify code during compliance checks — this skill assesses policy alignment only; any implementation must happen separately after compliance is confirmed.
2. ALWAYS run compliance check before HIGH/EPIC implementation — high-risk changes that bypass compliance checks create undetected policy drift that compounds over time into systemic violations.
3. ALWAYS report findings with specific remediation tasks and owning agent — vague "policy violation" reports without actionable remediation steps don't produce fixes; every FAIL and CONDITIONAL must include a concrete task.
4. NEVER report PASS on partial compliance — a plan that satisfies 80% of policies is a CONDITIONAL, not a PASS; partial compliance masks the remaining violations and gives false confidence.
5. ALWAYS recheck after remediation, not just once — a single compliance check before implementation is insufficient; verify again after major changes to confirm remediations are complete.

Workflow

Step 1: Gather Policy Context

• Read relevant files in .claude/rules/
• Read applicable workflow/agent constraints
• Read enforcement hook docs if needed

Step 2: Evaluate Proposed Change

Assess against:

1. Creator guard and artifact lifecycle rules
2. Routing and specialist-first requirements
3. Security and quality gate requirements
4. Memory/search/token-saver policy expectations

Step 3: Produce Decision

Return one policy decision:

• PASS: policy-aligned
• CONDITIONAL: allowed with required mitigations
• FAIL: not policy-compliant

Use this output shape:

{
  "decision": "PASS|CONDITIONAL|FAIL",
  "policyFindings"