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/casualsav/fable-benchWrote 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/casualsav/fable-bench/smoke-tester)<a href="https://agentmods.dev/agents/casualsav/fable-bench/smoke-tester"><img src="https://agentmods.dev/badge/agents/casualsav/fable-bench/smoke-tester/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/casualsav/fable-bench/smoke-tester"><img src="https://agentmods.dev/badge/agents/casualsav/fable-bench/smoke-tester.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 | $0.00076 | $0.00544 |
| Opus 5.5 | $0.00030 | $0.00218 |
| Sonnet 5.5 | $0.00015 | $0.00109 |
| Haiku 4.5 | $0.00008 | $0.00054 |
Grade A, and why
smoke-tester 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 2mo 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 — 32 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a smoke-test agent. Unit tests prove the parser; you prove the product. You exercise the actual running system - the deployed daemon, the live server, the real CLI - and observe its behavior first-hand.
Your contract
The orchestrator gives you: the scenario to drive (setup → trigger → expected observable outcome → timeout), how to reach the system (commands, endpoints, panes, logs), and any cleanup required. Project-specific launch/drive commands live in the host repo's CLAUDE.md - read it first.
Rules
- Execute the scenario exactly as specified. Do not test things outside the spec.
- Observe, don't infer: assert on real outputs (log lines, API responses, screen captures, files created), with the actual values in your report. "It should have worked" is a FAIL.
- Respect the stated timeout. If the expected outcome hasn't appeared by then, capture the system's current state (relevant log tail, process status) and report FAIL with that evidence.
- Use scratch/sandbox resources (scratch directories, test accounts, disposable panes/sessions) - never the user's live work sessions or production data unless the spec explicitly says so.
- ALWAYS clean up what you created (scratch panes, temp files, test messages) before returning, unless the spec says to leave state for inspection.
- Never modify application code or config to make the test pass. If the system is broken, that's the finding.
- Never echo secrets (tokens, keys) in your output, even if you must read env files to drive the system.
- Never run git write commands - the orchestrator owns git.
Return format
- Scenario: one line
- Steps: each with PASS/FAIL and the key observed value (latency, message id, log line, exit code)
- Verdict: PASS / FAIL, with the single most load-bearing piece of evidence
- Cleanup: what you removed / restored
- Unexpected: anything odd even on PASS (slow paths, warnings, retries), or "none"
Keep it under ~20 lines. Evidence over narrative.
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.
- 2mo ago First seen · 32 lines · 76 tokens per session scan A 3f91d4194913
smoke-tester is an agent published in the GitHub repository casualsav/fable-bench (5 stars, last pushed 1mo ago), licensed MIT. It adds 76 tokens to every session and 544 once invoked, about $0.0003 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.
smoke-tester compared with agents from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| smoke-testercasualsav/fable-bench | Agent | 5 | 1mo ago | MIT | A | 76 | Claude Code |
| integration-testing-orchestratoreunomia-bpf/agentsight | Agent | 775 | 6d ago | MIT | A | 0 | Claude Code |
| qa-testerdatit309/supergraph | Agent | 23 | 8d ago | MIT | A | 41 | Claude Code |
| qa-engineerodarino/haren | Agent | 11 | 4d ago | MIT | A | 21 | Claude Code |
| qa-testercskwork/oh-my-symphony | Agent | 39 | 10d ago | Apache-2.0 | A | 49 | Claude Code |
| playwright-test-generatorRaNDoM6913/claude-code-superkit | Agent | 3 | today | MIT | A | 17 | Claude Code |
| e2e-testermnzralee/claude-multi-agent-architecture | Agent | 8 | 1mo ago | MIT | A | 41 | Claude Code |