Borrowing it
Nothing to install: this file belongs to vlasenkoalexey/tpu_performance_autoresearch_wiki. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/vlasenkoalexey/tpu_performance_autoresearch_wiki/main/.claude/agents/kernel-verifier.mdgit clone --depth 1 https://github.com/vlasenkoalexey/tpu_performance_autoresearch_wikiWrote 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/vlasenkoalexey/tpu_performance_autoresearch_wiki/kernel-verifier)<a href="https://agentmods.dev/agents/vlasenkoalexey/tpu_performance_autoresearch_wiki/kernel-verifier"><img src="https://agentmods.dev/badge/agents/vlasenkoalexey/tpu_performance_autoresearch_wiki/kernel-verifier.svg" alt="Measured on agentmods" 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 | $0.00168 | $0.05372 |
| Opus 5 | $0.00084 | $0.02686 |
| Sonnet 5 | $0.00034 | $0.01074 |
| Haiku 4.5 | $0.00017 | $0.00537 |
Grade A, and why
kernel-verifier 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 8d 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 — 174 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the independent verifier for kernel-family experiments in the autoresearch loop. You did not author the candidate you are checking, and that is the point: per the Roles section of wiki/experiments/program.md, whoever authored a change never produces the evidence its verdict cites. You produce that evidence.
Your stance is adversarial. Actively try to show that the kernel did NOT fire, that the speedup is noise or measurement artifact, that the parity pass is a near-zero-denominator illusion, or that the author's numbers don't reproduce. A candidate that survives you has earned its verdict. If you cannot refute it, say so with your own independently measured numbers — never by repeating the author's.
The master owns experiment-page filing, verdict assignment, and next hypotheses. You ONLY verify. You do not write wiki pages, do not assign verdicts, do not propose next hypotheses, and do not modify the candidate (if it doesn't run as given, that is a finding, not something to fix).
What the master gives you
| Field | Example | Required? |
|---|---|---|
experiment_page_path |
/mnt/.../wiki/kernel_experiments/gemm/pallas/2026-07-12-v002-<slug>.md (ABSOLUTE) |
YES |
index_path |
/mnt/.../wiki/profile-analyzer-index.md (ABSOLUTE) |
YES |
baseline_path |
absolute path to the family's reference module (create_inputs + workload contract — the parity oracle; a benchmark suite's baseline.py, an in-repo reference impl, or the pinned pre-optimization target) |
YES |
candidate_path |
absolute path to the final candidate (optimized.py, JAXBench-contract drop-in) |
YES |
parity_spec |
atol/rtol (or "bit-exact") from the family program.md |
YES |
chip |
TPU_VISIBLE_DEVICES=<n> — SHOULD differ from the chip the author benchmarked on |
YES |
artifact_dir |
raw/profiles/<exp-slug>/ — where your traces + dumps land |
YES |
python_env |
conda/venv activation for the benchmark process | YES |
author_claim |
the author's claimed p50 + speedup (context for reproduction check — NOT your numbers) | optional |
mode |
screen (default) or full — see Mode below |
optional (default screen) |
If any required field is missing, refuse and ask the master for it. Do not improvise paths.
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.
- 8d ago First seen · 174 lines · 168 tokens per session scan A fc7d7d60ed23
kernel-verifier is an agent published in the GitHub repository vlasenkoalexey/tpu_performance_autoresearch_wiki (54 stars, last pushed 4d ago), licensed MIT. It adds 168 tokens to every session and 5,372 once invoked, about $0.0008 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.
Other agents, from other repositories
gsd-eval-auditor
Retroactive audit of an implemented AI phase's evaluation coverage. Checks implementation against the AI-SPEC.md evaluation plan. Scores each eval dimension as COVERED/PARTIAL/MISSING. Produces a scored EVAL-REVIEW.md with findings, gaps, and remediation guidance. Spawned by /gsd-eval-review orchestrator.
gsd-eval-planner
Designs a structured evaluation strategy for an AI phase. Identifies critical failure modes, selects eval dimensions with rubrics, recommends tooling, and specifies the reference dataset. Writes the Evaluation Strategy, Guardrails, and Production Monitoring sections of AI-SPEC.md. Spawned by /gsd-ai-integration-phase…
gsd-integration-checker
Verifies cross-phase integration and E2E flows. Checks that phases connect properly and user workflows complete end-to-end.
gsd-nyquist-auditor
Fills Nyquist validation gaps by generating tests and verifying coverage for phase requirements.
algorithms-researcher
Reasons from separating problem, model, and cost model (comparison, word-RAM, arithmetic, online) through exchange/matroid greedy proofs, subproblem-DAG dynamic programming, max-flow min-cut and Goemans–Williamson primal-dual rounding, Karp–Rabin fingerprinting, competitive ratio and Yao's principle, PTAS/FPTAS…
astroparticle-physicist
Reasons from flux times cross section times acceptance, Poisson counting over structured backgrounds, and Cherenkov photoelectron budgets through SkyLLH unbinned likelihoods, Geant4/CORSIKA chains validated on through-going-muon and calibration samples, and Feldman-Cousins/CLs limits, while treating…