debug-ml-inference

debug-ml-inference is a skill for Claude Code, Codex from DuqueOM/ML-MLOps-Portfolio. It costs 20 tokens per session (2,202 once invoked), scanned A, original, MIT.

A troubleshooting procedure for machine-learning predictions in production FastAPI services, including slow responses, incorrect results, and blocked asynchronous work.

In plain words
What is it for?
It helps inspect worker settings, memory-based scaling, model calls, explainability code, package versions, and unusually high model metrics, then suggests a fix or specific remediation plan.
Why use it?
It checks common causes in a fixed order and requires evidence from commands, helping distinguish the actual cause from guesses.

Skill for Claude CodeCodex

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 skills/duqueom/ml-mlops-portfolio/debug-ml-inference
Any agent
npx skills add DuqueOM/ML-MLOps-Portfolio --skill debug-ml-inference
Clone the repo
git clone --depth 1 https://github.com/DuqueOM/ML-MLOps-Portfolio

Made for: Claude Code, Codex.

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 debug-ml-inference

README.md
[![agentmods](https://agentmods.dev/badge/skills/duqueom/ml-mlops-portfolio/debug-ml-inference.svg)](https://agentmods.dev/skills/duqueom/ml-mlops-portfolio/debug-ml-inference)
Your own site
<a href="https://agentmods.dev/skills/duqueom/ml-mlops-portfolio/debug-ml-inference"><img src="https://agentmods.dev/badge/skills/duqueom/ml-mlops-portfolio/debug-ml-inference.svg" alt="Measured on agentmods" height="20"></a>
Per session 20 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,202 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. 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 $0.00020 $0.02202
Opus 5 $0.00010 $0.01101
Sonnet 5 $0.00004 $0.00440
Haiku 4.5 $0.00002 $0.00220

Measured 5d ago against content hash 3db7ac6f9e80, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

debug-ml-inference scanned grade A with 1 finding 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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

- Bash(curl:*)
.devin/skills/debug-ml-inference/SKILL.md · 203 lines

How it starts

The opening of the file, as written. The whole thing — 203 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Debug ML Inference

Systematically diagnose and fix ML inference issues in production FastAPI services.

Inputs

  • $service-name: Name of the ML service to debug (e.g., bankchurn)

Goal

Identify the root cause of the inference issue and either fix it or provide a specific remediation plan with commands. Every check must produce evidence (command + output).

Steps

1. Anti-Pattern Checklist (DO THIS FIRST)

Run this diagnostic before deep debugging — most inference issues match one of these patterns:

# Check Command Pass If
D-01 Multiple workers grep -rn "workers" $service-name/Dockerfile $service-name/k8s/ --workers absent or exactly 1
D-02 Memory HPA grep -n "memory" $service-name/k8s/base/*hpa* Empty output
D-03 Sync predict grep -rn "\.predict|predict_proba" $service-name/app/ Direct model calls only inside sync helpers delegated by run_in_executor
D-04 TreeExplainer grep -rn "TreeExplainer" $service-name/ None, or only in try/fallback
D-05 == pinning grep "==" $service-name/requirements.txt No ML packages with ==
D-06 Suspiciously high metric Check MLflow: primary > 0.99? Below 0.99
D-07 SHAP background Check background has both classes Both high/low probs
D-08 Uniform PSI bins grep -n "np.linspace|uniform" $service-name/src/*/monitoring/ Uses np.percentile
D-09 Missing heartbeat grep -n "heartbeat" $service-name/k8s/ monitoring/ Alert rule exists
D-10 tfstate in git git ls-files | grep tfstate Empty output
D-11 Model in Docker grep -n "COPY.*model|ADD.*model" $service-name/Dockerfile No matches
D-12 No quality gates grep -rn "quality_gate|should_promote" $service-name/src/ Gate logic exists
D-21/D-22 Blocking prediction logs grep -rn "log_prediction" $service-name/app/ Logging is fire-and-forget and errors are swallowed
D-23 Probe split grep -rn '"/health"|"/ready"' $service-name/app/ $service-name/k8s/ /health is liveness, /ready gates on model + warm-up
D-24 SHAP rebuild per request grep -rn "KernelExplainer" $service-name/app/ Built once during artifact load/warm-up, not inside endpoint

Read the full file on GitHub · 203 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 · 203 lines · 20 tokens per session scan A 3db7ac6f9e80

Subscribe to this mod's changes

debug-ml-inference is a skill published in the GitHub repository DuqueOM/ML-MLOps-Portfolio (5 stars, last pushed 4d ago), licensed MIT. It adds 20 tokens to every session and 2,202 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.