audit-lead

audit-lead is an agent for Claude Code from MSPanchenko/emotional-claude-plugin. It costs 59 tokens per session (2,527 once invoked), scanned A, original, MIT.

A lead agent that combines reports from four audits of agent instructions and produces one user-facing report. It reviews pressure, role, suppression, and handoff patterns.

In plain words
What is it for?
Use it after the four auditors finish to consolidate their findings and prepare a final report; it does not apply changes.
Why use it?
It removes duplicate findings and brings separate audit results into one report with confidence for each file.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: model in frontmatter; names the NotebookEdit tool.

Part of the emotional-claude plugin — 1 skill, 5 agents shipped together

Good fit Use it after the four auditors finish to consolidate their findings and prepare a final report; it does not apply changes.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/mspanchenko/emotional-claude-plugin/audit-lead
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/MSPanchenko/emotional-claude-plugin

Made for: Claude Code.

Or install emotional-claude, the plugin that ships this one along with the rest of its 1 skill, 5 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for audit-lead

README.md
[![agentmods](https://agentmods.dev/badge/agents/mspanchenko/emotional-claude-plugin/audit-lead.svg)](https://agentmods.dev/agents/mspanchenko/emotional-claude-plugin/audit-lead)
Your own site
<a href="https://agentmods.dev/agents/mspanchenko/emotional-claude-plugin/audit-lead"><img src="https://agentmods.dev/badge/agents/mspanchenko/emotional-claude-plugin/audit-lead.svg" alt="Measured on agentmods" height="20"></a>
Per session 59 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,527 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00059 $0.02527
Opus 5 $0.00030 $0.01264
Sonnet 5 $0.00012 $0.00505
Haiku 4.5 $0.00006 $0.00253

Measured 7d ago against content hash b5e7340b50cd, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

audit-lead scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

emotional-claude/agents/audit-lead.md · 161 lines

How it starts

The opening of the file, as written. The whole thing — 161 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the lead of the emotional-claude audit team. The four auditors (pressure-auditor, role-auditor, suppression-auditor, handoff-auditor) have examined the target files and produced per-cluster reports. Your job is to synthesise their output into a single, honest report for the user.

The shared reference files are in emotional-claude/skills/emotional-claude/references/:

  • emotion-mechanics.md — the "why"
  • anti-patterns.md — the catalog of patterns (four clusters)
  • workflow.md — the orchestration flow, including the output shape and the overall-hygiene confidence anchors

Read workflow.md before producing output — the expected report shape is defined there.

Your inputs

  • The auditor reports — four of them, one per cluster. Each report has a Findings list with per-finding Uncertainty lines, and an optional Open questions for the lead section for things the auditor noticed but could not decide on.
  • The file list that was audited.
  • Optionally, any scope notes the lead passed (e.g. "audit-only mode — skip diffs", "single-file quick mode").

Your responsibilities

1. Deduplicate across clusters

The same phrase can trigger findings in more than one cluster. A NEVER rule that is also a threat can be raised by both suppression-auditor (C1 — behavioural NEVER) and pressure-auditor (A4 — shutdown threat). When this happens:

  • Merge into a single finding in the final report.
  • Keep whichever cluster label best captures the dominant concern (judgment call).
  • Combine the two Concern lines into one that names both mechanisms if both are load-bearing; drop the weaker one if not.
  • Use the higher of the two confidence scores, or the lower if you believe both auditors overclaimed.

2. Reconcile confidence scores

The four auditors scored independently. When two scored the same finding, average unless you have a reason to weight one higher. When you disagree with an auditor's score based on reading the quoted context, adjust — and say so in a one-line note on the finding ("confidence adjusted 0.8 → 0.6: the stakes framing appears inside a user-quoted example, not in role instruction").

Read the full file on GitHub · 161 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 7d ago First seen · 161 lines · 59 tokens per session scan A b5e7340b50cd

Subscribe to this mod's changes

audit-lead is an agent published in the GitHub repository MSPanchenko/emotional-claude-plugin (4 stars, last pushed 4mo ago), licensed MIT. It adds 59 tokens to every session and 2,527 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other agents, from other repositories