audit-chat-log

Name: audit-chat-log
Availability: InStock
Author: smartmarbles

Audit chat logs and PROBE reports to reconstruct agent actions and generate verdict reports.

4|Updated Apr 2, 2026

Search Tags:#logs #verification #audit #transcripts #probe #lens #violations

Ready to run

npx skills add https://github.com/smartmarbles/helm --skill audit-chat-log

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: audit-chat-log
Source: https://github.com/smartmarbles/helm/tree/main/.github/skills/audit-chat-log
Command: npx skills add https://github.com/smartmarbles/helm --skill audit-chat-log

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Audit chat logs and PROBE reports to reconstruct agent actions, verify claims, detect behavioral violations, and generate standardized audit reports.

Core Features & Use Cases

Reconstruct agent call trees from chat logs and transcripts.
Detect behavioral violations such as impersonation, memory scope mismatches, and approval gate bypass.
Produce per-TC verdict tables and violation logs.
Verify PROBE report claims against log evidence to assess truthfulness.
Provide session-level anomaly summary and a final truthfulness summary.

Quick Start

Invoke the audit-chat-log skill with a provided chat log JSON and the corresponding PROBE report to generate a full audit.

Frequently Asked Questions about audit-chat-log

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

FAQPage Schema

How do I audit chat logs to verify agent actions and detect behavioral violations?▼

Auditing chat logs involves reconstructing agent call trees to verify claims and detect behavioral violations like impersonation or approval gate bypass. This skill processes chat log JSON and PROBE reports to generate standardized audit outputs including per-TC verdict tables and violation logs.

What is a PROBE report truthfulness summary and how does it verify log evidence?▼

A PROBE report truthfulness summary verifies claims by anchoring every verdict to exact log evidence. It assesses report accuracy by reconstructing agent actions from chat transcripts and documenting UNVERIFIABLE entries with full search paths for complete traceability.

How do I generate per-TC verdict tables and session-level anomaly summaries from chat transcripts?▼

Generating per-TC verdict tables and session-level anomaly summaries requires applying audit modes like P1 and P3 to chat logs. This reconstructs agent call trees, tracks violations, and produces standardized reports ensuring every verdict cites exact log evidence.

Can I use this skill to track memory scope mismatches and approval gate bypasses in LENS workflows?▼

Yes, this skill detects behavioral violations including memory scope mismatches and approval gate bypasses in LENS workflows. It applies session anomaly detection to chat logs and PROBE reports, producing violation logs and final truthfulness summaries with evidence anchoring.

What is the best way to document UNVERIFIABLE entries when reconstructing agent call trees from chat logs?▼

Documenting UNVERIFIABLE entries requires recording full search paths used during verification. This skill reconstructs agent actions from chat logs and PROBE reports, ensuring unverified claims are documented with complete evidence trails for transparent auditing.

What limitations exist when auditing chat logs with missing or incomplete PROBE reports?▼

When PROBE reports are missing or incomplete, auditing chat logs may produce UNVERIFIABLE entries. The skill documents these entries with full search paths, but cannot generate complete truthfulness summaries or per-TC verdict tables without sufficient log evidence to anchor every claim.