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.
npx agentmods add skills/orlando-japan/claude-code-setting/production-debuggingnpx skills add orlando-japan/claude-code-setting --skill production-debugginggit clone --depth 1 https://github.com/orlando-japan/claude-code-settingWrote 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.
[](https://agentmods.dev/skills/orlando-japan/claude-code-setting/production-debugging)<a href="https://agentmods.dev/skills/orlando-japan/claude-code-setting/production-debugging"><img src="https://agentmods.dev/badge/skills/orlando-japan/claude-code-setting/production-debugging.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00027 | $0.01147 |
| Opus 5 | $0.00014 | $0.00574 |
| Sonnet 5 | $0.00005 | $0.00229 |
| Haiku 4.5 | $0.00003 | $0.00115 |
Grade A, and why
production-debugging 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 3d 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.
How it starts
The opening of the file, as written. The whole thing — 107 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Production debugging
Rule zero: don't make the incident worse. Read-only operations first. Write operations only with deliberate intent.
The mindset shift
Local debugging: reproduce freely, restart freely, attach a debugger, print everything.
Production debugging: every action is potentially observable to users. You cannot:
- Restart to "try again."
- Add a
printand redeploy if the build takes 10 minutes. - Run an experiment that might page other people.
- Attach an interactive debugger that pauses the process.
You must get the answer you need from what's already there, with the minimum perturbation.
Information sources, in order of safety
1. Existing dashboards and metrics (completely safe)
Start here. 80% of "what's wrong" questions are answered by existing metrics.
- Error rate trend: when did it start, how fast is it growing?
- Latency distribution: is p99 blown while p50 is fine? Or both?
- Saturation: CPU, memory, queue depth, connections, thread pool.
- Recent deploys: correlate the start time with deploys.
- Recent config / flag changes: same.
2. Logs (safe, with care)
Query the log store for the transition. Not "show me errors from the last hour" (too much) — "show me errors in this service between the exact minute things changed."
- Group by error message: are you seeing one kind of error or many?
- Search for the request ID of a failing user action.
- Look at logs just before the errors started: what was the system doing?
Be careful with tail -f on prod files — can be slow and noisy. Prefer the log store's query interface.
3. Traces (safe, informative)
If you have distributed tracing, a failing request trace shows exactly where time went. Find a failing trace and follow it service to service.
4. Database reads (safe if read-only; dangerous if not)
SELECTqueries are fine. Be mindful of expensive queries on the primary — prefer a read replica.- Never
UPDATE,DELETE, orINSERTas part of debugging. Write operations are mutations; mutations are changes; changes during an incident are risky. EXPLAIN ANALYZEis usually fine on analytics queries but can cost on transactional ones. Be cautious.
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.
- 3d ago First seen · 107 lines · 27 tokens per session scan A 1bd8dacb7ab8
production-debugging is a skill published in the GitHub repository orlando-japan/claude-code-setting (2 stars, last pushed 3mo ago), licensed MIT. It adds 27 tokens to every session and 1,147 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.
Other skills, from other repositories
frontmcp-production-readiness
Pre-production audit, hardening, and go-live checklists for FrontMCP servers. Use before shipping to verify security hardening, performance, reliability, and observability, and for target-specific production checklists: Node server (Docker, graceful shutdown, Redis session scaling), Vercel and edge (cold-start…
langsmith-observability
LLM observability platform for tracing, evaluation, and monitoring. Use when debugging LLM applications, evaluating model outputs against datasets, monitoring production systems, or building systematic testing pipelines for AI applications.
production-readiness
Use this skill to verify that an AI system is ready for production before deployment. Covers model abstraction, cost/latency, retries, fallbacks, structured logging, tracing, observability, evaluation in CI/CD, monitoring, and rollback planning. Activates as a mandatory gate before production deployment of any AI/LLM…
agent-observability
Instrument an LLM agent so failures are diagnosable, traces are replayable, and evals can run against production data. Use when the user is moving an agent past prototype and mentions tracing, spans, OpenTelemetry, LangSmith, Langfuse, Arize, OpenLLMetry, structured logs, GenAI semantic conventions, or asks "how do I…
ai-native-production-ops
AI Native 产品方法论——生产运行与循环回灌的实操 Skill。 用户提供产品已上线或即将上线，Skill 自动执行生产运行流程： 可观测性设计 → 指标体系搭建 → 失败与人工修正识别 → 反馈回灌 → 价值/商业/客户循环 → 输出生产运行方案。 包含价值发现循环（§1）、AIOps案例（§2）、客服案例（§3）、SaaS案例（§4）、产品团队（§5）五个合并子模块。 基于《AI Native 产品方法论》第18-27章。.
langsmith-observability
LLM observability platform for tracing, evaluation, and monitoring. Use when debugging LLM applications, evaluating model outputs against datasets, monitoring production systems, or building systematic testing pipelines for AI applications.