verify-paired-native-measurements
Skill4 starsChanged 22 days agoMITView on GitHub
Audit the process, sample and archive evidence behind paired native benchmarks, especially wrappers around an unchanged official client. Use before admitting benchmark results or when reviewing the controller that produced them. Ordinary unit tests do not need this.
In this repository79 other files it gives its agents.
- AGENTS.mdAGENTS.md
- accepted-proof-exact-retuneSkill
- arxiv-portable-researchSkill
- auto-autoresearch-pipelineSkill
- beachhead-ratchetSkill
- bounded-commuting-foldSkill
- burn-the-house-downSkill
- calibrated-static-cost-proxySkill
- circuit-edit-integration-safetySkill
- colab-laneSkill
- computer-algebra-oraclesSkill
- constraint-construction-searchSkill
- cross-model-review-gateSkill
- deepseek-pressure-testSkill
- ecdsafail-fastevalSkill
- ecdsafail-lowq-playbookSkill
- ecdsafail-oracle-loopSkill
- ecdsafail-pareto-ratchetSkill
- ecdsafail-scratch-dietSkill
- elicit-researchSkill
- evaluator-scoring-surface-auditSkill
- exact-evaluator-replicaSkill
- exact-hold-passSkill
- exact-postpass-stackSkill
- exact-search-engineeringSkill
- fast-lean-proof-loopSkill
- fit-probe-and-spliceSkill
- gpu-affine-table-synthesisSkill
- greedyfrog-methodSkill
- hard-problem-contractSkill
- hold-demand-ledgerSkill
- hosted-carry-lendingSkill
- inert-op-screen-to-structural-ruleSkill
- invariant-chartSkill
- jev-second-perspectiveSkill
- lambda-exchange-deskSkill
- leader-rebase-stackSkill
- lean-arklib-proofsSkill
- lean-finite-certificate-splittingSkill
- lever-switches-and-break-controlsSkill
- mission-commandSkill
- nasqret-playbookSkill
- paged-walk-descentSkill
- peak-lifetime-ledgerSkill
- pinned-table-stackingSkill
- pip-disciplineSkill
- portable-source-assertionsSkill
- proximity-prize-campaignSkill
- proximity-prize-frontier-archaeologySkill
- quantum-affine-lookup-synthesisSkill
- quantum-borrowed-wire-counterexampleSkill
- quantum-dense-qromSkill
- quantum-dirty-borrowingSkill
- quantum-divstep-reconstructionSkill
- quantum-encoded-arithmeticSkill
- quantum-graded-divider-placementSkill
- quantum-guarded-permutation-auditSkill
- quantum-mbundo-lifetimesSkill
- quantum-record-miningSkill
- quantum-replay-arithmeticSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 21
Use focused counterexamples at actual admission boundaries.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Zwillingswaag (verify-paired-native-measurements)
Bind the conclusion to the declared experiment, exact native artifacts and actual retained evidence. Local comparisons do not establish an official score or promotion. This skill adds no execution, provider, spending or submission authority. Use the existing mission and its canonical Mission Command policy when applicable.
Admission boundaries
- Preserve the official client's ordering, sample counts, timing endpoints and statistic definitions when that is the experiment's contract. Disclose wrapper overhead included inside those endpoints; do not silently subtract it or claim equivalence to the bare client.
- Enforce elapsed limits with a monotonic clock and separately enforce absolute work/grant deadlines. Carry remaining limits across process boundaries. A wall-clock rollback must not authorize a new subprocess after the elapsed cap. Include preparation, verification and finalization in whichever caps the declared design assigns them.
- Across machines, monotonic timestamps have unrelated origins. Keep the controller's original budget start and bind a separate worker clock/boot observation. Transfer a conservatively reduced remaining duration; never compare the raw clocks or restart the allowance at worker launch. A declared end-to-end cap also covers the final local readback. See cross-host budgets when designing that boundary.
- Validate every retained complete run, including warmups. Exclusion from statistics does not waive positive time/RSS, correctness, byte identity or other declared evidence requirements.
- Reconstruct the exact conditional phase progression. A passing prefix with missing successor phases is incomplete. A completed early stop requires the declared failed gate and corresponding terminal decision. Recompute the gate from rows rather than trusting its stored label.
- Retain the direct parent's actual child wait status and raw streams. Worker success or process disappearance cannot replace an unavailable supervisor exit. Bind PID/start identity when process IDs may be reused.
- Distinguish pre-archive snapshots from final terminal sidecars. Validate the implementation's actual status transition instead of either accepting a stale success label or requiring fields saved at different phases to be identical.
- Snapshot the exact expected receipt bytes before use. Verify the compressed archive, every declared member, and the fresh restore. Recheck the same compressed identity after semantic validation and at successful publication. Emit the identity actually validated. A newly computed unchecked hash is not provenance.
Use focused counterexamples at actual admission boundaries. Observed failure cases give four useful examples; adapt them to the declared experiment rather than importing its historical counts or thresholds.
Preserve failed evidence. Missing exits, truncated archives, timeouts and incomplete phases close admission; they do not by themselves refute the candidate algorithm.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

