Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› issta-writing-style
issta-writing-style
GitHub指导ISSTA论文修订，强调明确测试贡献、威胁模型先行、评估契约化及去匿名化。规范声明范围与证据配对，撰写实质性有效性威胁分析，并提供句式重写示例，确保符合18页ACM格式及专业审稿要求。
Trigger Scenarios
ISSTA论文修改
软件测试论文写作
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill issta-writing-style -g -y
SKILL.md
Frontmatter
{
"name": "issta-writing-style",
"description": "Use when revising an ISSTA paper for a clear testing\/analysis contribution, covering the threat-model-then-technique structure, the evaluation contract, a threats-to-validity section that is not boilerplate, claims scoped to the subjects tested, double-anonymous phrasing, and 18-page ACM sigconf discipline that survives specialist reviewers."
}
ISSTA Writing Style
Use this when revising the main paper. An ISSTA paper has to state, early and precisely, what software behaviour it finds or reasons about, on what it was evaluated, and what the evidence showed — to a reviewer who already knows the nearest tool and benchmark.
Revision rules
- Put the contribution on the first page: the problem in a concrete testing/analysis setting, why existing techniques miss it, the technique, and what the evaluation showed. Name the technique; do not hide it behind "a novel approach."
- State the threat model or analysis scope before the technique. Reviewers judge soundness against the scope you set, so an unstated scope is an unbounded claim they will puncture.
- Make the evaluation a contract: subjects, baselines, metric, and statistic, declared before the numbers. "Outperforms prior work" without that contract is the single most common weak sentence.
- Pair every claim with evidence: a table, a benchmark result, an artifact path, or a scoped limitation. A declarative sentence near a result is something a specialist reviewer will check.
- Write a real threats-to-validity section: internal (confounds in the measurement), external (do the subjects generalize), and construct (does the metric measure the property). A boilerplate threats paragraph is noticed and counts against presentation.
- Keep double-anonymous phrasing in self-citations, tool names that de-anonymize, repository references, and acknowledgements.
Contribution-presentation discipline
- Scope claims to the subjects tested. "Detects order-dependent flaky tests" is defensible; "detects flaky tests" when you tested one subclass is not.
- Separate what the technique guarantees from what it empirically achieves; mixing a soundness claim with a benchmark number in one sentence is a credibility leak.
- Define the analysis or measurement once, in one place; two-column ACM pages punish redefinition.
- When a result is best-effort or unsound-but-useful, say so; ISSTA reviewers respect a scoped, honest technique over an overclaimed "sound" one.
Sentence-level rewrites
| Draft pattern | ISSTA-safe rewrite |
|---|---|
| "Our tool significantly outperforms prior work" | "finds X% more true bugs than |
| "We detect flaky tests" | "We detect the order-dependent subclass of flaky tests" |
| "Evaluated on real-world programs" | "Evaluated on <benchmark vN> across N subjects (SHAs in the artifact)" |
| "Our analysis is sound" | "Sound under Assumptions 1-2; native calls are out of scope" |
Vignette: compressing into 18 ACM pages
A draft with a formal analysis, six subjects' worth of tables, and a sprawling related-work section: keep the analysis rules, one soundness sketch, the evaluation protocol, and the two decision-critical tables in the body; compress related work into delta-first contrasts; push complete per-subject tables and extra configurations into the artifact. The test of a good cut: a specialist reviewer can judge soundness and evaluation without opening the artifact — the artifact deepens the paper, it does not complete it.
What a strong ISSTA opening does
1. names the concrete testing/analysis problem and its cost
2. bounds the scope (which subclass, which language, which assumption)
3. names the technique and its one key idea
4. states the evaluation contract: subjects, baseline, metric, statistic
5. reports the finding, honestly scoped
Output format
[Writing diagnosis] clear / under-scoped / overclaimed / contract-missing
[First-page fix] <new framing>
[Scope statement] <threat model / analysis scope>
[Claim discipline] <claim -> table/artifact/limitation>
[Compression cuts] <move to artifact / delete / merge>
[Anonymity edits] <phrases to rewrite>
Version History
- 9f86f09 Current 2026-07-19 16:06
