[Feature]: Preset strategy to replace one section of a template
维护者通常 1 天内回复
还没有人认领这个 Issue。
评估
调研方向
Start in PresetResolver.resolve_content (the composition loop) and VALID_PRESET_STRATEGIES, then read manifest validation for the template strategies. The issue also touches the bash, PowerShell and Python runtime scripts and the preset docs (developing.md, api.md). Agree the design with maintainers before coding, since the author's fork is not a PR yet. Done means the acceptance criteria pass: heading-scoped replacement, exactly-one-match failure, fence-aware matching, and validation errors for commands and scripts.
由索引模型根据 Issue 内容生成。
描述
Problem Statement
A preset can change a template in four ways today: replace, prepend, append and wrap. None of them can change one section in the middle of a template. To change only ## User Scenarios & Testing in spec-template.md, a preset has to replace the whole file. That copy then stops picking up upstream changes to every other section, and it stops composing with other presets that touch the same template.
Follow-up to discussion #5326, where @mnriem asked for the cases and a fuller write-up.
Proposed Solution
A fifth template strategy, replace-section, with a section field naming the heading:
provides:
templates:
- type: "template"
name: "spec-template"
file: "templates/outcomes.md"
strategy: "replace-section"
section: "User Scenarios & Testing *(mandatory)*"
The file holds the new body of that section. The rules:
- Find the section by its heading text. Match the text after the
#marks exactly, ignoring surrounding whitespace. Any heading level counts. - Keep the heading line. Replace only the body, from after the heading to the next heading of the same or higher level, or the end of the file.
- Exactly one match, or fail. If no heading matches, or more than one does, the preset fails with an error that lists the template's headings. A renamed or duplicated section fails loudly instead of being appended or guessed at.
- Ignore headings inside fenced code blocks.
- Compose like the other strategies. It runs on the resolved lower-priority content, the same base
wrapuses, so two presets can each replace a different section of the same template. Likeappend,prependandwrap, it needs a lowerreplacebase. - Templates only at first. Commands and scripts keep the strategies they have.
In the code, this looks like one more branch in the composition loop of PresetResolver.resolve_content, replace-section added to VALID_PRESET_STRATEGIES, and manifest validation requiring section when the strategy is replace-section.
Alternatives Considered
replacethe whole template. This works today, but the copy drifts from upstream and blocks other presets from composing on that template.- Marker comments in the core templates (
<!-- section:stories -->). They're more precise, but every core template would need markers, and a project's hand-edited template would lose them. Headings already exist, and they're what people and models use to find their way around a template. wrapwith a hand-written copy of the other sections. Same drift problem asreplace.
Component
Spec templates (BDD, Testing Strategy, etc.)
Use Cases
These come from SpecKit Companion, which uses this rule in its own template renderer:
- Spec: replace
## User Scenarios & Testingwith observable outcomes (no P1/P2/P3 stories) for small changes, or with EARS-style numbered requirements (WHEN/THEN/SHALL) that tasks can cite by ID. - Plan: a team's own
## Technical Contextblock with its stack's fields, and every other section stock. - Tasks: replace
## Notesto rule out tasks an agent can't do (UAT, deploy, comms), or the format section so every task says how to verify itself.
Acceptance Criteria
- A preset with
strategy: "replace-section"andsection: "<heading>"replaces only that section's body in the resolved template, and keeps the heading line. - Every other section of the template comes out the same as without the preset.
- The section ends at the next heading of the same or higher level, so its subsections are replaced with it.
- Zero or more than one matching heading fails with an error that names the section and lists the template's headings.
- Headings inside fenced code blocks are not matched.
- Two presets replacing different sections of one template both apply, in priority order.
- Manifest validation rejects
replace-sectionwithoutsection, and rejects it on commands and scripts. - The preset docs (
developing.md,api.md) describe the strategy.
Additional Context
A working version is on my fork, not opened as a PR: full diff against main (branch feat/replace-section-strategy). It covers the CLI resolver, manifest validation, the bash/PowerShell/Python runtime scripts (so feature creation gives the same output as the CLI), the artifact listings, docs and tests. I'm happy to open it as a PR if this direction is agreed.
SpecKit Companion's version of this is replace_section in apps/speckit-extension/scripts/template_render.py in alfredoperez/speckit-companion, with tests. It currently takes the first of two matching headings. The "exactly one match" rule above is stricter, and I'd change ours to match.
AI Disclosure
- Issue text: drafted with Claude Code (Claude Opus 5.5, default settings, interactive under my supervision) after reading the spec-kit preset resolver and docs. I reviewed and edited it before posting.
- Linked branch: implemented with Claude Code (Claude Opus 5.5, default settings) in agent mode running autonomously, following a spec I wrote and reviewed. Covers code generation, tests and docs. It's a proof of concept for this discussion, not a PR.
- 主要语言
- Python
- 星标
- 149k
- 派生
- 13.4k
- 平均合并
- 1 天 5 小时
- 30 天内合并 PR
- 223
环境准备
在浏览器里用你自己的 GitHub 账号启动这个项目的开发容器。
- 没有 Dockerfile 或 Docker Compose 文件
- 有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
github/spec-kit 的其他 Issue
-
难度 2/5 1-3 小时 新手友好度 68/100
维护者通常 1 天内回复
-
triage-must-have
难度 1/5 1 小时以内 新手友好度 78/100
维护者通常 1 天内回复
-
Workflow-step verifier test fixture breaks when the checkout path contains whitespace (4 failures on macOS)可能已有人在做 @mehdikarimi55 于 3 天前认领。 未关闭triage-can-wait
难度 2/5 1-3 小时 新手友好度 83/100
维护者通常 1 天内回复
-
Parity tests fail under the documented direct pytest invocation when the ambient python3 lacks PyYAML (11 failures)可能已有人在做 @mehdikarimi55 于 3 天前认领。 未关闭triage-must-have
难度 2/5 1-3 小时 新手友好度 82/100
维护者通常 1 天内回复
-
bug triage-out-of-scope
难度 2/5 1-3 小时 新手友好度 64/100
维护者通常 1 天内回复
相似的 Issue
-
bug
难度 2/5 1-3 小时 新手友好度 80/100
glossyio/traffic-monitor#312 ·
维护者通常 1 天内回复
-
documentation enhancement
难度 2/5 1-3 小时 新手友好度 76/100
es-ude/elastic-ai.creator#1046 ·
维护者通常 3 天内回复
-
feature request
难度 2/5 1-3 小时 新手友好度 65/100
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 72/100
confident-ai/deepteam#306 ·
维护者通常 1 天内回复
-
SslLibrary crashes with unhandled OSError: port/proto not found on non-standard SSL ports可能已有人在做 @himanchalkaushale 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 82/100
OWASP/Nettacker#1884 · 1 条评论 ·
维护者通常 1 天内回复