Use when drafting ASE (IEEE/ACM Automated Software Engineering) author responses, covering the double-anonymous rebuttal that must first survive the early-rejection gate and — distinctively — the criteria-bound revision-round summary-of-changes that maps every stated revision criterion to a concrete change and is re-checked against those criteria.
SkillScore95Maintained
Automatic
◆Research pack
Last verified 8/15/2026
What it changes
Passed 20 of 21 test cases with the skill, and 10 of 21 without it.
Without skill45%
With skill95%
Checks it passed that the model alone missed
Identifies the speaking turn as rebuttal.
Categorizes baseline fairness doubt under soundness.
Categorizes model-versus-design doubt under soundness.
and 11 more in the full report
11 fixed10 passed either way1 still failing0 broke
Pass-rate lift
+50 pp
Tokens
+34%
Agent turns
0%
Response time
-30%
Model gemini-3.6-flashMeasured 8/15/2026v121 of 22 cases passed with skill · 21 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 10
Prompt
Reviewer 2 stated: 'Section 3.2 is incomprehensible; equations 3 and 4 use un-defined symbols N_t and K_s, making the formal model impossible to follow.' Which decision dimension does this concern belong to, and what status action is appropriate?
What was checked, without and with the skill
Checked
Without skill
With skill
Categorizes un-defined notation under clarity.
Passed
Passed
Marks the resolution status as DONE upon adding symbol definitions in the text.
Missed
Passed
case-20 · gemini-3.6-flash · 8/15/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md88 linessingle file
ASE Author Response
Use this after ASE reviews are released. ASE has two distinct speaking turns, and conflating
them is a common mistake: a short rebuttal (for papers that survive the early-rejection gate),
and — if you receive a Revision — a summary-of-changes accompanying a revised paper that is
re-checked against the reviewers' explicitly stated criteria. Both must respect double-anonymity:
no author names, institutions, or revealed repository ownership, even to strengthen a point.
Triage (both turns)
Answer what affects the decision: significance and soundness of the automation, evaluation quality
on real subjects, threats to validity, clarity, and the Data Availability / artifact posture.
Use evidence that already exists or that the revision will concretely add — never a vague promise.
Correct factual misreadings first; a reviewer who misread your tool's configuration or a table is
often persuadable.
Keep every word anonymous. Do not name your tool's real name, institution, grant, or a private
repository.
Turn 1 — the rebuttal (only if you survived early rejection)
If your paper reached the rebuttal stage, at least one reviewer saw value; a uniformly negative set
would have been early-rejected. So the rebuttal's job is narrow: move the borderline reviewer.
Short and decision-focused. One decision-critical point per reviewer beats an exhaustive reply.
Concede what is true, correct what is misread, and point to exactly where the answer lives in the
submitted paper or artifact.
Do not paste large new results the reviewers cannot verify; if a small, checkable number answers a
direct question ("what is the runtime overhead?"), give it.
Signal what you would do in a revision, because Revision is a live outcome — but do not promise a
redesign you cannot deliver in the window.
Turn 2 — the revision summary-of-changes (the distinctive ASE move)
A Revision is criteria-bound: the reviewers wrote concrete, actionable criteria and agreed to
accept in principle if they are met. The summary-of-changes is a criterion ledger: for every
stated criterion, either meet it and show where, or explain precisely why it is infeasible and what
you did instead.
[C1] Criterion (quoted): "Add a comparison against tool X on the same subjects."