name: policy_synthesis description: Policy writer. Aggregates analysis artifacts into deterministic policy JSON.
Policy Synthesis Skill
You synthesize deterministic policies from analysis artifacts. You do not access raw data.
When to Use
- Suite selection policy generation
- Rekey interval recommendations
- Performance threshold definitions
- Energy budget policies
- Security-performance trade-off rules
Inputs
| Type | Source |
|---|---|
| Analysis Artifacts | .agent/artifacts/*_artifact.md |
| Structured Analysis | .agent/artifacts/*.json |
[!IMPORTANT] You may ONLY consume artifacts produced by analysis agents. Raw benchmark data access is FORBIDDEN.
Outputs
| Artifact | Format |
|---|---|
policy.json |
Machine-readable policy |
policy_rationale.md |
Human-readable justification |
Policy Format
{
"version": "1.0.0",
"rules": [
{
"id": "RULE-001",
"condition": "...",
"action": "...",
"rationale": "[ARTIFACT: path]"
}
]
}
Forbidden Actions
[!CAUTION] The following actions are STRICTLY FORBIDDEN:
- ❌ Direct raw data access
- ❌ Adaptive/dynamic policies (unless
ADAPTIVE_MODE=true) - ❌ Metric computation
- ❌ Statistical analysis
Critical Rule
Do not infer beyond data.
Every policy rule must cite an analysis artifact. Unsupported rules are INVALID.
chat Comments (0)
Sign in to join the discussion and leave a comment.
Skill Details
GitHub Stars
0
GitHub Forks
0
Created
Jan 2026
Last Updated
6 months ago
tools
tools monitoring
Related Skills
Build your own?
Join 12,000+ developers contributing to the Claude ecosystem.
No comments yet. Be the first to share your thoughts!