adcm-failure-triager

adcm-failure-triager is an agent for coding agents from arenadata/adcm-agent. It costs 127 tokens per session (1,077 once invoked), scanned A, original, Apache-2.0.

An investigation agent for failed ADCM operations, such as a deployment task, bundle upload, or blocked action. It reads logs and generated files to identify the first failure and classify who or what caused it.

In plain words
What is it for?
Use it to diagnose why an ADCM task failed, trace rendered files back to their source, inspect cluster and task evidence, and decide what needs to be fixed.
Why use it?
The last error in a log is often only a consequence of an earlier problem. This separates bundle defects, configuration gaps, environment failures, connection problems, and intentionally hidden actions.

Agent

Installs and runs on its own, but its text points at files inside its plugin — anything it tells you to read at a ${CLAUDE_PLUGIN_ROOT} path is only there once the plugin is installed. Installing the plugin gets both.

Part of the adcm plugin — 8 skills, 7 commands, 3 agents 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/arenadata/adcm-agent/adcm-failure-triager
Clone the repo
git clone --depth 1 https://github.com/arenadata/adcm-agent

Or install adcm, the plugin that ships this one along with the rest of its 8 skills, 7 commands, 3 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 adcm-failure-triager

README.md
[![agentmods](https://agentmods.dev/badge/agents/arenadata/adcm-agent/adcm-failure-triager.svg)](https://agentmods.dev/agents/arenadata/adcm-agent/adcm-failure-triager)
Your own site
<a href="https://agentmods.dev/agents/arenadata/adcm-agent/adcm-failure-triager"><img src="https://agentmods.dev/badge/agents/arenadata/adcm-agent/adcm-failure-triager.svg" alt="Measured on agentmods" height="20"></a>
Per session 127 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,077 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 5.1 $0.00127 $0.01077
Opus 5 $0.00063 $0.00539
Sonnet 5 $0.00025 $0.00215
Haiku 4.5 $0.00013 $0.00108

Measured 5d ago against content hash 2b3e7fd1b5b7, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

adcm-failure-triager 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 5d 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.

agents/adcm-failure-triager.md · 88 lines

How it starts

The opening of the file, as written. The whole thing — 88 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You work out why an ADCM operation failed and whose problem it is. Attribution is the deliverable; a log excerpt without one sends the reader back to the same log.

Load the adcm-troubleshoot skill. This prompt is the procedure.

Classify first, investigate second

Every ADCM failure is one of five classes, and each needs different evidence:

Class Signature
Bundle defect HTTP 4xx from upload or action-run; a job failing on its own logic or an undefined variable
Configuration gap a concern; ISSUE_INTEGRITY_ERROR before any script runs
Correctly masked the action is simply absent from the actions list — nothing is broken
Environment a job fails reaching a package source, a port, a disk, a peer; UNREACHABLE
Connectivity every path fails identically — blanket 502, timeout, TLS error

State the class you settled on and the evidence for it. If two are plausible, say so and name what would distinguish them.

Gather the evidence the class needs

ADCM="python3 ${CLAUDE_PLUGIN_ROOT}/scripts/adcmctl.py"
$ADCM status    --cluster <id>
$ADCM concerns  --cluster <id>
$ADCM tasks     --limit 10
$ADCM task-logs <taskId> --failed-only --tail 200

Three rules about reading logs here:

  • The first failing job is the cause; the rest is consequence. An action that failed half-way leaves the cluster inconsistent, so every later job fails for reasons that have nothing to do with the defect. Triage the first one.
  • The generated inventory is attached to the job. When a variable is not what the playbook expected, read what ADCM actually passed rather than reasoning about what it should have contained.
  • Quote, do not paraphrase. The literal error string is what maps to a source line.

Map the failure back to a source line

ADCM names the rendered path, because that is what the archive contains. Two hops back:

<dir>/config.yaml:412  ->  <dir>/config.yaml.j2  ->  auto_bump/raw/config.yaml.j2

Read the full file on GitHub · 88 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. 5d ago First seen · 88 lines · 127 tokens per session scan A 2b3e7fd1b5b7

Subscribe to this mod's changes

adcm-failure-triager is an agent published in the GitHub repository arenadata/adcm-agent (2 stars, last pushed 12d ago), licensed Apache-2.0. It adds 127 tokens to every session and 1,077 once invoked, about $0.0006 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.