fizz-debug

A debugging guide for FizzBee specifications, which describe system behavior for checking possible states and actions. It helps investigate failed checks, unexpected results, slow runs, and incorrect state spaces.

In plain words
What is it for?
Use it to read failure traces, replay guided sequences, inspect available actions, visualize smaller state spaces, and diagnose timeouts or incorrect state counts.
Why use it?
It provides a structured way to understand why a verification run failed or explored the wrong states. It helps reproduce failures, inspect action traces, reduce the problem, and investigate performance issues.

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

Made for: Claude Code, Codex.

Per session 49 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,634 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 $0.00049 $0.01634
Opus 5 $0.00024 $0.00817
Sonnet 5 $0.00010 $0.00327
Haiku 4.5 $0.00005 $0.00163

Measured 2d ago against content hash c12806daae8f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

fizz-debug 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 2d 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.

.claude/skills/fizz-debug/SKILL.md · 209 lines

How it starts

The opening of the file, as written. The whole thing — 209 lines — stays where its author put it; the contents beside it link to each section on GitHub.

FizzBee Debugging Guide

Full references: ~/.claude/skills/fizzbee-docs/VERIFICATION_GUIDE.md, ~/.claude/skills/fizzbee-docs/GOTCHAS.md, ~/.claude/skills/fizzbee-docs/PERFORMANCE_GUIDE.md


When Model Checker Reports FAILED

A FAILED result means a safety assertion was violated. The output includes a counterexample trace.

Step 1: Read the failing trace

The output shows the sequence of actions that led to the violation. Look at:

  • Which assertion failed (AlwaysFoo, BalanceNonNegative, etc.)
  • What state variables were when it failed
  • What action just ran before the failure

Step 2: Reproduce with a guided trace

Copy the failing action sequence and replay it:

fizz --trace "Node#0.RequestLock
Node#1.RequestLock
Node#0.DoWork" spec.fizz

Add --trace-extend 1 to see what enabled actions exist at each step:

fizz --trace "Node#0.RequestLock" --trace-extend 1 spec.fizz

Step 3: Reduce config to minimum

Shrink the state space so you can visualize it:

fizz --preinit-hook "N=1" spec.fizz
dot -Tsvg out/*/graph.dot -o graph.svg && open graph.svg
# Graph auto-generated when < 250 nodes

In the SVG, look for:

  • The state where the assertion fails (highlighted)
  • What changed in the step before failure

When Trace Is Incomplete

WARNING: Trace execution incomplete. Expected 8 links, executed 6 links.

A transition in your trace was blocked. Most common causes:

  1. A require guard failed — a condition you expected to be true wasn't
  2. A symmetry limit was exceeded (e.g., limit=2 but trace creates 3 distinct values)
  3. The action name in the trace doesn't match (case-sensitive, exact match)

Debug by running the trace up to the failing step and using --trace-extend 1 to see what's actually enabled:

fizz --trace "Step1
Step2" --trace-extend 1 spec.fizz

When Behavior Seems Wrong (But PASSED)

The model may be over-constraining (silent pruning) or assertions may be tautological.

Read the full file on GitHub · 209 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. 2d ago First seen · 209 lines · 49 tokens per session scan A c12806daae8f

Subscribe to this mod's changes

fizz-debug is a skill published in the GitHub repository fizzbee-io/fizzbee (347 stars, last pushed 7d ago), licensed Apache-2.0. It adds 49 tokens to every session and 1,634 once invoked, about $0.0002 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 2026-08-30.

Related

Other skills, from other repositories

tla-check

Write and iteratively refine executable TLA+ specs (.tla) and TLC model configs (.cfg) from natural-language system designs; run TLC model checking; summarize pass/fail and counterexamples with explicit assumptions and bounds. Use when asked to design or validate a protocol/state machine, create or edit .tla/.cfg…

younes-io/agent-skills · 79 tokens

tla-proof

Write and iteratively refine TLA+ theorem proofs in .tla modules with TLAPS (tlapm); run proof checks and summarize proved vs failed/omitted obligations with explicit assumptions and trust boundaries. Use when asked to create or fix THEOREM or PROOF blocks, diagnose TLAPS failures, strengthen inductive invariants…

younes-io/agent-skills · 87 tokens

vera-language

Write programs in the Vera programming language. Use when asked to write, edit, debug, or review Vera code (.vera files). Vera is a statically typed, purely functional language with algebraic effects, mandatory contracts, and typed slot references (@T.n) instead of variable names.

aallan/vera · 61 tokens

allium

Give your AI agents something more useful than a prompt. Velocity through clarity.

juxt/allium · 18 tokens

elicit

Run a structured discovery session to build an Allium specification through conversation. Use when the user wants to create a new spec from scratch, elicit or gather requirements, capture domain behaviour, specify a feature or system, define what a system should do, or is describing functionality and needs help…

juxt/allium · 66 tokens

propagate

Generate tests from Allium specifications. Use when the user wants to propagate tests, generate test files from a spec, write tests for a specification, create property-based tests, produce state machine tests, check test coverage against spec obligations, or understand what tests a specification requires.

juxt/allium · 57 tokens