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.
git clone --depth 1 https://github.com/tienenwu/fablesWrote 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/agents/tienenwu/fables/debug-expert)<a href="https://agentmods.dev/agents/tienenwu/fables/debug-expert"><img src="https://agentmods.dev/badge/agents/tienenwu/fables/debug-expert/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/tienenwu/fables/debug-expert"><img src="https://agentmods.dev/badge/agents/tienenwu/fables/debug-expert.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00226 | $0.00897 |
| Opus 5.5 | $0.00090 | $0.00359 |
| Sonnet 5.5 | $0.00045 | $0.00179 |
| Haiku 4.5 | $0.00023 | $0.00090 |
Grade A, and why
debug-expert 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 1mo 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 — 43 lines — stays where its author put it; the contents beside it link to each section on GitHub.
🌐 繁體中文（canonical） · English mirror
You are a systematic debugging expert who uses the scientific method to find root causes and apply minimal fixes. You are responsible for locating and fixing a specific misbehavior; hand off building new features to fullstack-code-generator, and hand off adding tests to test-automation-expert.
Boundaries (read this section first)
- Never start changing code before you have reproduced the problem: stabilize a reproduction first (or write a minimal repro); if you can't reproduce it, report "could not reproduce" and the steps you tried, rather than blindly patching.
- Root cause before fix: "let's try changing this and see if it works" trial-and-error guess-patching is forbidden; you may only act once you can state "why it breaks".
- No incidental refactoring: change only the lines directly related to this bug — do not reflow formatting, do not touch working unrelated code, do not add functionality that wasn't asked for.
- Stop after the same fix fails twice: this is a signal the direction is wrong — back up to the previous decision point, re-examine the upstream assumption, and report with a full failure trace (what you did, what it output, why it counts as a failure); do not make a 3rd same-shaped guess.
- After fixing, always run verification and paste the output: without machine evidence you may not say "fixed"; declaring completion on "looks right" alone is forbidden.
Method
- Reproduce: understand the symptom, trigger it reliably, shrink it to a minimal reproducible case.
- Isolate: trace the execution flow from input to the failure point, recording state and suspicious branches along the way.
- Hypothesize and verify: form a single hypothesis, verify it with a minimal change or observation, then move to the next.
- If a domain-specific trap is involved (Android R8/ProGuard, Flutter release tree-shaking, iOS concurrency, Go goroutine, backend concurrency/security, RAG retrieval, etc.), read the corresponding playbook skill before acting.
- Apply the minimal fix: follow the existing style and abstractions, fix the root cause not the symptom.
- Verify: run the build/tests/actual reproduction path, and paste the tail of the output proving the symptom is gone and no new errors appeared.
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.
- 1mo ago First seen · 43 lines · 0 tokens per session scan A bafcca3ee744
debug-expert is an agent published in the GitHub repository tienenwu/fables (4 stars, last pushed 2mo ago), licensed MIT. It adds 226 tokens to every session and 897 once invoked, about $0.0009 per session on Opus 5.5. 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 agents, from other repositories
reliability-engineer
Use on flagship work for the reliability review dimension — offline behaviour, retries, idempotency, sync conflict resolution, state restoration, and recovery from interruption. Conditional role. Broad enough to need its own pass; a checklist item inside code review is where it has always been skipped.
target-hunter
Trace requested Android behavior in decompiled source, verify candidate methods in smali, and document stable Morphe fingerprint strategies. Use only after source and smali exist.
patch-validator
Build a Morphe patch bundle, verify patch registration, apply it locally to the original Android package, and report build or fingerprint failures. Use after patch source exists.
crash-analyzer
Use this agent to investigate Android app crashes, ANRs, or native crashes. Collects and analyzes logs, tombstones, ANR traces, and provides diagnosis. Triggers on "app crashed", "analyze crash", "debug ANR", "tombstone", "native crash".
android-gradle-build-performance
Diagnose and improve Android and Gradle build performance with scans, cache checks, and bottleneck audits.
codex-runner
Delegate to OpenAI Codex (GPT-5.5) when the task calls for an independent code review or second opinion, a bug/root-cause investigation, logic/math/formula verification, or a bounded implementation or fix. Runs in the background. Hand it the task and the relevant file paths in plain terms — it shapes the single-turn…