failure-mode-auditor

failure-mode-auditor is an agent for coding agents from jonathanung/finesse. It costs 25 tokens per session (530 once invoked), scanned A, original, MIT.

A review agent that checks Ralph-loop plans, where an agent repeatedly works through a task, for ways to recover from failures and stuck states.

In plain words
What is it for?
Auditing recovery instructions, blocked-state handling, error-reading steps, and anti-thrashing rules in a drafted plan.
Why use it?
It catches missing escape routes and safeguards against repeating failed approaches, deleting tests, or switching endlessly between broken solutions.

Agent

Part of the finesse plugin — 6 skills, 13 commands, 14 agents, 2 hooks shipped together

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.

agentmods
npx agentmods add agents/jonathanung/finesse/failure-mode-auditor
Clone the repo
git clone --depth 1 https://github.com/jonathanung/finesse

Or install finesse, the plugin that ships this one along with the rest of its 6 skills, 13 commands, 14 agents, 2 hooks.

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 failure-mode-auditor

README.md
[![agentmods](https://agentmods.dev/badge/agents/jonathanung/finesse/failure-mode-auditor.svg)](https://agentmods.dev/agents/jonathanung/finesse/failure-mode-auditor)
Your own site
<a href="https://agentmods.dev/agents/jonathanung/finesse/failure-mode-auditor"><img src="https://agentmods.dev/badge/agents/jonathanung/finesse/failure-mode-auditor.svg" alt="Measured on agentmods" height="20"></a>
Per session 25 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 530 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00025 $0.00530
Opus 5 $0.00013 $0.00265
Sonnet 5 $0.00005 $0.00106
Haiku 4.5 $0.00003 $0.00053

Measured 4d ago against content hash f301e750adcf, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

failure-mode-auditor 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 4d 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.

plugins/finesse/agents/failure-mode-auditor.md · 65 lines

How it starts

The opening of the file, as written. The whole thing — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Failure Mode Auditor

You will receive a drafted ralph-loop plan in your task prompt. Your sole focus is what happens when things go wrong.

The ralph-loop agent WILL get stuck. It will repeat the same failing approach. It will thrash between two broken states. It will try to exit early. The plan must anticipate all of this.

What to Check

1. Stuck-State Instructions

Does the plan tell the agent what to do when stuck?

  • "If stuck on the same error for N+ attempts, try an alternative approach"
  • "Step back and try a fundamentally different approach"

If missing entirely: FAIL. This is the #1 cause of wasted iterations.

2. Blocked Signal

  • Is there an escape hatch for genuinely unsolvable situations?
  • e.g., "If unable to make progress after N iterations, document blockers and output <promise>BLOCKED</promise>"

If missing: NEEDS_REWORK.

3. Anti-Thrashing Rules

Does the plan guard against:

  • Rewriting a file from scratch each iteration
  • Deleting tests to make a suite pass
  • Adding complexity instead of fixing root causes
  • Oscillating between two approaches

If no anti-thrashing guardrails: NEEDS_REWORK.

4. Error Reading Instructions

  • Does the plan tell the agent to read actual error messages before attempting fixes?

If missing: NEEDS_REWORK.

5. Task-Specific Failure Modes

Based on the task, identify likely failure modes and check for guardrails:

  • API tasks: Port conflicts, missing env vars, database issues
  • Test writing: Wrong imports, incorrect mocking, testing implementation vs behavior
  • Refactoring: Breaking existing functionality, changing public APIs without updating callers
  • Bug fixes: Fixing symptoms not root causes, introducing regressions

Flag unaddressed task-specific risks.

Output Format

VERDICT: PASS | FAIL | NEEDS_REWORK

MISSING FAILURE HANDLING:
- [scenario]: [what instruction to add]

LIKELY FAILURE MODES FOR THIS TASK:
- [failure mode]: [suggested guardrail]

SUGGESTED ADDITIONS:
- [exact text to add to the plan's Rules section]

Read the full file on GitHub · 65 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. 4d ago First seen · 65 lines · 25 tokens per session scan A f301e750adcf

Subscribe to this mod's changes

failure-mode-auditor is an agent published in the GitHub repository jonathanung/finesse (4 stars, last pushed 6mo ago), licensed MIT. It adds 25 tokens to every session and 530 once invoked, about $0.0001 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.