agentleFS
darrenCWJdarrenCWJ/agent-skills-scaffold/.claude/skills/debugging/SKILL.md

debugging

Skill0 starsChanged 3 months agoMITView on GitHub

Hypothesis-driven debugging for any failing system — code, pipeline, document build, tool call, or process. Use whenever something errors, behaves unexpectedly, produces wrong output, or fails intermittently, and ALWAYS before attempting a second blind fix ("let me just try changing X" twice in a row means this skill should be active).

121 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use ssr, Problem-Solving Techniques and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • JetBrainsSkill

    ssr

    Guide for using Structural Search and Replace in IntelliJ. Use when creating or modifying SSR inspections or search patterns.

    Also in 3 projects that use this

  • truongtv22Skill

    Problem-Solving Techniques

    Apply systematic problem-solving techniques for complexity spirals (simplification cascades), innovation blocks (collision-zone thinking), recurring patterns (meta-pattern recognition), assumption constraints (inversion exercise), scale uncertainty (scale game), and dispatch when stuck. Techniques derived from Microsoft Amplifier project patterns adapted for immediate application.

    Also in 4 projects that use this

  • mrgoonieSkill

    media-processing

    Process multimedia files with FFmpeg (video/audio encoding, conversion, streaming, filtering, hardware acceleration) and ImageMagick (image manipulation, format conversion, batch processing, effects, composition). Use when converting media formats, encoding videos with specific codecs (H.264, H.265, VP9), resizing/cropping images, extracting audio from video, applying filters and effects, optimizing file sizes, creating streaming manifests (HLS/DASH), generating thumbnails, batch processing images, creating composite images, or implementing media processing pipelines. Supports 100+ formats, hardware acceleration (NVENC, QSV), and complex filtergraphs.

    Also in 6 projects that use this

  • mrgoonieSkill

    web-frameworks

    Build modern full-stack web applications with Next.js (App Router, Server Components, RSC, PPR, SSR, SSG, ISR), Turborepo (monorepo management, task pipelines, remote caching, parallel execution), and RemixIcon (3100+ SVG icons in outlined/filled styles). Use when creating React applications, implementing server-side rendering, setting up monorepos with multiple packages, optimizing build performance and caching strategies, adding icon libraries, managing shared dependencies, or working with TypeScript full-stack projects.

    Also in 6 projects that use this

  • DevNewbie1826Skill

    remove-ai-slops

    Remove AI-generated code slop from branch changes or an explicit file list while preserving behavior, including junk tests (prose pins, source greps, copied inventories, mock echoes), each proven by its class rule before a cut. Use when asked to remove slop, clean AI code, deslop, clean up generated code, strip slop, prune junk tests, or review generated-code cleanup. Start by locking behavior with regression tests, then use the workflow and category references.

    Also in 3 projects that use this

  • DevNewbie1826Skill

    programming

    MUST USE for work on Python, Rust, TypeScript, JavaScript, or Go source files and project manifests, and whenever a test or spec file is written, reshaped, reviewed, or deleted. Triggers: write or edit code; add, change, or delete a test; test authoring gate; junk tests; new project; Go HTTP service (gin lane); Bubble Tea TUI; CJK IME; Connect-Go RPC; sqlc or pgx; branded IDs; exhaustive matching; unsafe Rust; Miri; oversized file; refactoring; logging, log levels, structured logging, or observability; arena, allocator, bumpalo, const fn, const generics, compile-time programming, zero-allocation APIs, bitfields, repr, scope guards, errdefer-style cleanup, Zig-like patterns, zero-copy parsing, packed data, or FFI; Rust public API design or macros.

    Also in 3 projects that use this

  • +6 more from mrgoonie

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 109 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Keep a running list in the task notes: H1: <hypothesis> → <experiment> → CONFIRMED/REFUTED (<evidence>).

    line 22

“Common” is measured: kept by at least 6 in 10 of the 109 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Debugging

Debugging is science, not thrashing. One observation, one hypothesis, one experiment at a time. Two consecutive blind fixes = stop and start this protocol properly.

The loop

  1. Reproduce first. Find the minimal, deterministic reproduction. If you cannot reproduce it, you cannot verify any fix — reproduction IS the first deliverable. For intermittent failures, find the frequency ("fails ~1 in 5") so you know how many clean runs prove a fix.
  2. Read the actual error. The full message, the full stack, the actual line. Do not pattern-match on the first word and guess. The answer is in the error more often than pride allows.
  3. State the expected vs observed in one sentence each. If you can't state expected behavior precisely, the bug might be in your understanding of the spec.
  4. Form ONE hypothesis that explains ALL observations — not just the most recent one. A hypothesis that ignores an observation is wrong.
  5. Design the cheapest discriminating experiment — one that will give a different result depending on whether the hypothesis is true. Prefer observation (logging, inspecting state, bisecting input) over mutation (changing code to see what happens).
  6. Run it. Update. Hypothesis confirmed → fix the ROOT CAUSE, not the symptom. Refuted → record it (see ledger below) and return to step 4 with the new information.
  7. Verify the fix (verification skill): reproduction now passes, AND the rest still passes — a fix that breaks something else is a new bug, not a fix.

The hypothesis ledger

Keep a running list in the task notes: H1: <hypothesis> → <experiment> → CONFIRMED/REFUTED (<evidence>). This prevents re-testing dead hypotheses, makes handovers survivable mid-debug, and exposes when you're going in circles.

Standard moves when stuck

  • Bisect — halve the input / commit range / pipeline until the failure isolates. Fastest generic weapon.
  • Diff against last-known-good — what changed between working and broken? (git diff, file timestamps, dependency versions, environment).
  • Check the boring things — wrong file, stale cache, wrong directory, wrong environment, unsaved change, case-sensitive path. Rule these out EARLY; they are disproportionately common and maximally embarrassing late.
  • Question the diagnosis, not just the fix — after 3 refuted hypotheses, the frame is probably wrong: the bug may be in a different layer entirely.

Exit rules

Debugging respects the loops skill budget. After the budget (default 3 full loop iterations) is spent: write up the reproduction, the hypothesis ledger, and the narrowed search space — then escalate to the user. That write-up has real value; thrashing does not.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.