aaai-artifact-evaluation — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited aaai-artifact-evaluation (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Use this to prepare artifacts that reviewers can use to assess reproducibility. AAAI supplementary material is part of the submission record; after review starts, do not assume it can be updated.
experiments.
commands, seeds, expected outputs, and runtime.
rebuttal.
AAAI does not run a separate badged artifact-evaluation committee the way some systems venues do; the same broad-AI reviewer who scores the paper also inspects whatever supplement you attach. That reviewer may be a planning, knowledge-representation, or constraint-satisfaction specialist rather than a deep-learning engineer, so the artifact has to be legible without insider tooling. Optimize for a reviewer who skims, not one who will spend an afternoon configuring a cluster.
| Reviewer action | Passes | Fails |
|---|---|---|
| Opens the ZIP | sane tree, top README | nested archives, 0-byte files |
| Reads appendix | maps to numbered claims | contradicts the paper |
| Tries one command | reproduces one headline number | needs private data or credentials |
| Scans for identity | nothing reveals authors | Git logs or home paths leak |
Because clearly-below-bar papers can be cut before author feedback, a supplement that looks thin or unrunnable is a cheap reason to summary-reject. Avoid these:
A constraint-solving paper claims a 30% node-expansion reduction. The team ships a large ZIP of raw solver logs but no driver script. The reproduction path is empty, so artifact status is "risky"; the fix is a small run_main.py that regenerates Table 2 from seeds, a trimmed log sample, and a license for the benchmark instances. The raw dump moves to the post-acceptance release.
[Artifact status] complete / partial / risky / unavailable
[Submitted files] technical appendix / multimedia appendix / code-data ZIP
[Reviewer reproduction path] <commands and expected output>
[Anonymity risks] <metadata, links, paths, logs>
[Missing items] <data, code, seeds, licenses, hardware>~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.