debug

A diagnostic method for finding the root cause of a software bug without applying the fix. It uses a failing reproduction, testable hypotheses, focused experiments, and a record of rejected explanations.

In plain words
What is it for?
Use it to reproduce a failure, investigate its cause, try temporary diagnostic probes, and hand the confirmed root cause and regression test to the implementation work.
Why use it?
It replaces random edits with evidence-based investigation. Keeping the fix for a separate task preserves review, testing, and regression coverage.

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/progiri/craftlight/debug
Any agent
npx skills add progiri/craftlight --skill debug
Clone the repo
git clone --depth 1 https://github.com/progiri/craftlight

Made for: Claude Code, Codex.

Per session 244 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,642 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 $0.00244 $0.02642
Opus 5 $0.00122 $0.01321
Sonnet 5 $0.00049 $0.00528
Haiku 4.5 $0.00024 $0.00264

Measured yesterday against content hash c4b730503fd9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

debug 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 yesterday.

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.

1. **Build the feedback loop.** The repro is one command — a test, a curl, a CLI run over a fixture — that
plugins/craftlight/skills/debug/SKILL.md · 133 lines

How it starts

The opening of the file, as written. The whole thing — 133 lines — stays where its author put it; the contents beside it link to each section on GitHub.

debug — the systematic root-cause hunt

Principle: debug sits BELOW task — a diagnostic subcycle that task drops into by the no-guess-and-patch stop rule and returns from with a root cause. Observation over guessing: a hypothesis must yield a testable prediction before the experiment, and every rejected option is recorded. Blind loops of "tweaked at random — reran" are the main token-eater; a systematic pass is cheaper than three guesses.

The boundary is fix-shaped, not edit-shaped. debug doesn't write the fix and doesn't commit one — not even an obvious one-liner; the root found → the fix goes through task (classification, the gate, a regression test). But diagnostic edits are the skill's own tools and are allowed: temporary log lines, probe tests, a narrow assertion. Two rules keep them honest: instrumentation is never committed and is stripped before the diagnosis is handed off — tag every probe with one unique prefix (e.g. DBG-7f3), so the strip is a single grep and no untagged stray survives; the reproduction test from step 1 is the exception — it is left uncommitted and passed to task as the ready-made regression test (task's fix begins red on it). The discipline's symmetry: plan plans and stops, code-review reads and doesn't edit, debug diagnoses and doesn't fix.

Not for this skill: the cause is known, only the fix remains → task; "how does X work" → a normal answer; assessing code quality → code-review. The cause became obvious from the very first error → don't unfold the ceremony: return the diagnosis in one phrase and hand off to task.

Step 0. Orientation and resume

The craftlight block upsert (reference and procedure — skills/task/templates/CLAUDE-block.md) rides with the creation of DEBUG.md — the only file debug writes. A direct entry whose log stays in the chat touches the repo not at all, so it must not edit the user's CLAUDE.md; an entry from task finds the block already ensured by task.

There's a docs/crafts/*/DEBUG.md with status in-progress:

  • A bare resume trigger ("back to the bug", "resume debugging") → show the last hypothesis and current status from the log and offer to continue: DEBUG is external memory, resuming reads the log, not the chat history. Several in-progress DEBUGs → ask which one.
  • A new bug in the message → take it on as usual; no blocking question — one FYI line: parked: debug "<slug>" is in-progress — say "back to the bug" to resume it.

Read the full file on GitHub · 133 lines

Files

What ships with it

2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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. yesterday First seen · 133 lines · 244 tokens per session scan A c4b730503fd9

Subscribe to this mod's changes

debug is a skill published in the GitHub repository progiri/craftlight (5 stars, last pushed 7d ago), licensed MIT. It adds 244 tokens to every session and 2,642 once invoked, about $0.0012 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.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

next-cache-components-adoption

Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…

vercel/next.js · 95 tokens

babysit-pr

Babysit a GitHub pull request after creation by continuously polling review comments, CI checks/workflow runs, and mergeability state until the PR is merged/closed or user help is required. Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and keep…

openai/codex · 114 tokens

imagegen

Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output…

openai/codex · 113 tokens

cpu-profile-analysis

Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…

microsoft/vscode · 71 tokens

next-cache-components-optimizer

Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…

vercel/next.js · 170 tokens