| name | limitations-failure-case-auditor |
| description | Audits limitations, assumptions, failure cases, negative results, boundary conditions, and ethical or deployment risks in Robotics and AI papers. Use when the user needs a limitations section, failure analysis, responsible research check, or pre-review overclaim audit. |
Limitations Failure Case Auditor
Use this before final paper review, rebuttal, or camera-ready submission.
Inputs
- Draft paper, abstract, introduction, experiments, conclusion, or limitations section.
- Benchmark results, failed runs, qualitative examples, logs, safety notes, and deployment constraints.
- Target venue or responsible research checklist if known.
- Desired depth:
quick, standard, or deep; default to quick for daily overclaim review.
Read references/limitations-checks.md for standard or deep audits, or when claims involve safety, humans, deployment, or responsible research risk.
Workflow
For quick triage, inspect only claims whose scope is likely too broad: generality, robustness, safety, deployment, data efficiency, real-world transfer, and negative-result omissions.
- Extract claims that imply generality, safety, robustness, deployment, fairness, or real-world readiness.
- Identify assumptions: data, robot platform, sensors, simulator, tasks, metrics, compute, user population, and operating conditions.
- Map observed failures and missing evaluations to the claims they weaken.
- Separate limitations from excuses: a limitation must tell readers how to interpret the result.
- Draft limitation and failure-case language that is specific, honest, and not self-sabotaging.
- Recommend minimal additional analysis when a limitation can be quantified.
- Flag claims that must be narrowed before review.
Output
For quick triage:
## Claims to Narrow
| Claim | Why scope is too broad | Safer wording |
| --- | --- | --- |
## Missing Failure Evidence
For standard or deep audits:
## Limitations and Failure Case Audit
## Assumption Map
| Claim | Assumption | Evidence | Risk |
| --- | --- | --- | --- |
## Failure Cases
| Failure | Where observed | Interpretation | Paper action |
| --- | --- | --- | --- |
## Suggested Limitations Text
## Claims to Narrow
## Additional Analysis Needed
Rules
- Do not hide major limitations in vague future-work language.
- Do not invent failure cases that were not observed or plausibly implied by evidence.
- Do not over-apologize for normal scope boundaries.
- Do not make safety, robustness, or deployment claims without matching evaluation.
- Do not treat limitations as optional if the venue explicitly asks for them.
Validation
- In quick triage, each claim-to-narrow has a concrete scope reason and safer wording.
- In standard or deep audits, each limitation connects to a claim, assumption, result, or missing evaluation.
- Failure cases explain what readers should infer.
- Strong claims are narrowed when evidence is absent.
- Responsible research concerns are checked when humans, robots, safety, or societal deployment are in scope.
Completion
Done means the chosen depth is explicit, scope boundaries are visible, failure cases are interpretable when checked, and overclaims are either supported or narrowed.