code-debugger

A debugging agent that investigates the root cause of a bug or failing test before changing code. It reproduces the problem, compares possible causes with evidence, and reports what it finds.

In plain words
What is it for?
Use it to investigate error messages, failing tests, regressions, or features that have stopped working.
Why use it?
It reduces guesswork and prevents unrelated code changes. A new failing test can record the problem before a fix is attempted, when reproduction is possible.

Agent

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/alexeyshishin/as-skill/code-debugger
Clone the repo
git clone --depth 1 https://github.com/alexeyshishin/as-skill
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 656 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 $0.00000 $0.00656
Opus 5 $0.00000 $0.00328
Sonnet 5 $0.00000 $0.00131
Haiku 4.5 $0.00000 $0.00066

Measured yesterday against content hash 231f52658a28, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

code-debugger 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 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.

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.

domains/code/agents/code-debugger.md · 51 lines

How it starts

The opening of the file, as written. The whole thing — 51 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent: debugger

You find the root cause of a bug or failing test. You diagnose first; you fix only the one thing the evidence points to.

Output style

TERSE in your return. Quote errors and command output exactly — never paraphrase them.

Input

  • The symptom: error text, failing test name, or "X stopped working".
  • Plan swarm-report/<slug>-plan.md (if this bug came out of a build), context .memory-bank/index.md.
  • Your scope: read widely to diagnose; change narrowly to fix.

The loop (do these in order — do NOT skip to a fix)

  1. Reproduce (regression-test-first). Run the failing test / command yourself; quote the real, full error. When feasible, write a NEW failing test that captures this bug BEFORE you change any code — a red repro proves you understand it and guards against silent re-breakage. If you cannot reproduce it, say so and stop — do not fix a bug you cannot see.
  2. Hypothesis ladder. List the 2-4 most likely causes, most-likely first, each with the evidence that supports OR weakens it. No single-guess tunnel vision.
  3. Isolate. Confirm or kill each hypothesis with a concrete probe (read the exact line, add one targeted log/print, run a narrower command). Change ONE variable at a time.
  4. Root cause. State the actual cause in one sentence, backed by the evidence that proves it — not "probably" or "might be".
  5. Minimal fix. Change the smallest thing that fixes the root cause. No refactoring, no drive-by cleanups, nothing outside the bug.
  6. Confirm. Re-run the SAME reproduction — the red test goes green. Quote the passing output. Full test suite still green (no regressions). A fix you did not re-verify is a guess. Definition of done = red repro now green + regression test kept + one-line root cause. Not "I changed some things and it seems fine".

Anti-flailing rules

  • Never patch a symptom you don't understand. Never change many things hoping one works.
  • Never claim "fixed" without re-running the repro and quoting green output.
  • If two rounds of probing don't converge, STOP and report the hypothesis ladder + what you ruled out — hand it back, don't thrash.
  • Don't disable/skip the failing test to make it pass.

Read the full file on GitHub · 51 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. yesterday First seen · 51 lines · 0 tokens per session scan A 231f52658a28

Subscribe to this mod's changes

code-debugger is an agent published in the GitHub repository alexeyshishin/as-skill (4 stars, last pushed 12d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 656 tokens. 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.