asae

Runs iterative convergence-gate audits of output artifacts against sources and specifications.

Updated Apr 11, 2026
One-click install
npx skills add https://github.com/gozonerd/shadow-ai-assessment --skill asae-gozonerd
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: asae
Source: https://github.com/gozonerd/shadow-ai-assessment/tree/main/.claude/skills/asae
Command: npx skills add https://github.com/gozonerd/shadow-ai-assessment --skill asae-gozonerd

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Output artifacts (documents, code, designs) often drift from their original sources and specifications, and single-pass self-review misses systematic errors. ASAE enforces a repeatable, threshold-based audit loop that only exits after a configurable number of consecutive clean passes, verified by an independent rater. ## Core Features & Use Cases - Iterative Convergence Gate: Repeats the same comprehensive audit pass until the configured certainty threshold of consecutive clean passes is reached, or halts with escalation after max iterations. - Severity-Classified Findings: Classifies every finding as CRITICAL, HIGH, MEDIUM, or LOW, with standard or strict policies controlling counter resets and exit conditions. - Domain Audit Checklists: Applies tailored checklists for document, code, design, research, instructional design, legal, and other domains, including security, accessibility, and compliance criteria. - Independent Rater Verification: Spawns a separate subagent with zero shared context to confirm the audit verdict before issuing PASS, producing a versioned audit log as reproducibility evidence. - Use Case: A parent workflow like /dare-to-rise-code-plan invokes ASAE at every stage boundary to gate code or document deliverables before they are treated as complete. ## Quick Start Run an ASAE convergence gate on my output file against its source materials and the original spec with a certainty threshold of 3.

Frequently Asked Questions about asae

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a convergence gate audit on an output artifact?

Invoke /asae with a scope definition containing the target artifact paths, source material paths, the original prompt or spec, a domain, a certainty threshold, and a severity policy. The gate iterates full audit passes until the threshold of consecutive clean passes is reached.

What inputs does the ASAE audit require?

ASAE requires a scope definition with target, sources, prompt, domain, asae_certainty_threshold, and severity_policy fields. An optional max_iterations field defaults to 10 and triggers a halt with escalation when exceeded.

What is the difference between standard and strict severity policy?

Under the standard policy, MEDIUM findings do not reset the clean-pass counter but must be fixed before exit. Under the strict policy, MEDIUM findings reset the counter to zero, which suits high-stakes outputs like regulatory filings or production code.

Why does ASAE require an independent rater before passing?

Single-persona audits have structural blind spots because the auditor often authored the artifact. ASAE spawns a separate subagent with no shared context to re-evaluate the same checklist, and only a CONFIRMED verdict allows the audit to pass.

What happens when the audit exceeds the maximum iterations?

The gate halts and returns a HALT status to the caller with the final pass findings. The parent skill or user then decides whether to escalate, re-scope the audit, or abandon the artifact.