WEB VALIDATOR · NO ACCOUNT REQUIRED

Paste a robot repository. Find the reproducibility gaps.

Inspect a public GitHub or Hugging Face repository before attempting reproduction. KnownRobot pins the source revision, reads approved metadata only, and turns missing declarations into clear next steps. Configuration review never proves physical transfer or safe operation.

policy / terminal

$ robot-skill check ./policy --strict

INCOMPLETE  ./policy
Detected 4 evidence files
ERROR  policy.framework_version
  Pin the framework version.
ERROR  hardware.gripper
  Declare the end effector.
WARN   evaluations
  Compatibility remains unverified.

INTERACTIVE VALIDATOR

Paste a repository. See what is reproducible.

KnownRobot resolves the revision, inspects an approved metadata allowlist, and explains what was found, what is missing, and what still requires human review. No account required.

CLI / DEEPER VALIDATION

Need local files, CI, or a portable manifest?

1
Installpipx install . # from the Known Robot 1.6.0 checkout
2
Inspect a policyrobot-skill check ./policy
3
Use it in CIrobot-skill check . --strict --no-write

Python 3.10–3.12 · macOS, Linux, and Windows · Apache-2.0

02 / EVIDENCE CONTRACT

One file. Eight evidence groups.

The manifest stays beside the policy on GitHub or Hugging Face. Null values are allowed structurally so the first scan can be saved; strict mode fails until every required claim is supplied.

01Policy

Framework, pinned version, architecture, checkpoint

02Hardware

Robot family, gripper, sensors, calibration references

03Runtime

Control frequency, observation and action shapes

04Dataset

Repository, revision, feature schema, license

05Dependencies

Resolved package versions and hashed declarations

06Source

Repository type, immutable revision, artifact version

07Compatibility

Supported, unsupported, and untested hardware

08Evaluation

Benchmark, trials, successes, evaluator, evidence

03 / RELIABLE AUTOMATION

Designed to fail usefully.

Exit 0

The scan ran. In default mode, an incomplete draft is still written so contributors can improve it.

Exit 1

The path, manifest, or filesystem could not be read. No misleading output is written.

Exit 2

Invalid evidence always fails. Strict or complete validation also fails on missing metadata; target comparison fails on mismatches or unknowns.

JSON diagnostics

--format json returns stable finding codes, paths, severities, and a summary for CI.

Use --target ./target.yaml --no-write --format json to compare declared configurations. Equal shapes alone are not sufficient: feature semantics and calibration matter. A match never establishes safe physical transfer or independent reproduction.

04 / MEASURED SIMULATION

Run the policy. Preserve every outcome.

The separate local evaluator executes actual policy inference in FetchPickAndPlace-v4 or FetchReach-v4 using Gymnasium Robotics and MuJoCo. It records seeded resets, observations, actions, final simulator success signals, failures and inference latency. This is Fetch simulation evidence—not SO-101 physical compatibility or safety certification.

python -m pip install '.[evaluation]'
robot-skill evaluate evaluation.yaml --allow-execution --output evidence/run-01
robot-skill verify-evaluation evidence/run-01

Use a real Fetch manifest and actual policy artifact. Your YAML/JSON configuration declares format: knownrobot-evaluation/1.0, environment, planned trials, initial seed, max_steps, trial_timeout_seconds, a relative manifest path and a policy object containing kind, relative path and the actual file sha256.

Policies may be NumPy MLP archives or reviewed local Python factories. Python execution additionally requires --trust-policy: arbitrary policy code is not sandboxed. Use an isolated environment without secrets or attached hardware. Metadata validation still never executes code.

Bundles include the complete trial ledger, execution traces, input hashes, package versions and checksum inventory. Failed trials are retained. Integrity checks do not establish independent verification. Add --evidence-url with your chosen HTTPS bundle location to generate an eligible submission.json; publish the complete evidence in your existing repository and review attribution before submission.

Complete evaluation configuration, policy interface and publication guide ↗ · Submit measured evidence →

05 / EXTERNAL ASSESSMENT

Pin, inspect and publish without executing.

The admin workflow resolves or verifies a full Hugging Face commit, downloads only allowlisted metadata, inserts verified repository provenance into the generated manifest, hashes every inspected file and emits a publication-ready assessment bundle.

robot-skill assess-hf aadarshram/act_pusht \
  --revision 6d403b142934aaef61fc07f5eec1515c4325751f \
  --artifact-intent task_policy --artifact-intent simulation_policy \
  --output assessments/aadarshram-act-pusht

Artifact intent is descriptive only; it does not change completeness or establish compatibility. Add --claims upstream-claims.json for reviewed, categorized model-card paraphrases. After verification, sync-assessment advances the record through draft, review and published database states; export-assessments creates the source-controlled audit snapshot. Checkpoint weights and policy code are never downloaded or executed.

06 / DURABLE REPORT

Keep the findings with the manifest.

robot-skill check ./policy --format assessment --output assessment.json
robot-skill verify-report assessment.json --policy ./policy

The report includes every diagnostic, the validator version, inspected-file hashes, provenance, declared target comparison and an explicit metadata-only execution boundary. Atomic writes do not replace existing reports unless --force is supplied.

STANDARDIZE IN THE OPEN

Use the format without joining a platform.

The schema, example, implementation, and tests are public. Keep the manifest with your policy and review it like code.

Read the CLI guide ↗Download schema 1.0 ↗Reusable GitHub Action and portable credit guide ↗Inspect adapter ownership →Check LeRobot dataset-version compatibility →Corrections and disputes →Report a transfer failure →

Loading published evidence…

$ robot-skill check ./policy --strict INCOMPLETE ./policy Detected 4 evidence files ERROR policy.framework_version Pin the framework version. ERROR hardware.gripper Declare the end effector. WARN evaluations Compatibility remains unverified.

Designed to fail usefully.

Exit 0

The scan ran. In default mode, an incomplete draft is still written so contributors can improve it.

Exit 1

The path, manifest, or filesystem could not be read. No misleading output is written.

Exit 2

Invalid evidence always fails. Strict or complete validation also fails on missing metadata; target comparison fails on mismatches or unknowns.

JSON diagnostics

--format json returns stable finding codes, paths, severities, and a summary for CI.

Run the policy. Preserve every outcome.

python -m pip install '.[evaluation]' robot-skill evaluate evaluation.yaml --allow-execution --output evidence/run-01 robot-skill verify-evaluation evidence/run-01

Pin, inspect and publish without executing.

robot-skill assess-hf aadarshram/act_pusht \ --revision 6d403b142934aaef61fc07f5eec1515c4325751f \ --artifact-intent task_policy --artifact-intent simulation_policy \ --output assessments/aadarshram-act-pusht

Keep the findings with the manifest.

robot-skill check ./policy --format assessment --output assessment.json robot-skill verify-report assessment.json --policy ./policy