Add an AGENTS.md rule for pacing bulk GitHub writes (secondary rate limit)
维护者通常 1 天内回复
还没有人认领这个 Issue。
评估
- 难度
- 2/5
- 预计耗时
- 1-3 小时
- 新手友好度
- 76/100
- Issue 类型
- 文档
- 描述清晰度
- 描述清楚
- 活跃度
- 活跃
- 技术栈
- github, typescript
- 领域
- devtools
调研方向
从 AGENTS.md 开始，查找有关 issue 驱动工作和长期工作的指南，然后阅读 docs/contribution-model.md 中的 Close 步骤，以及用于批量写入 GitHub 的 issue-triage 和 board-ops 技能操作指南。将提供的规则添加到 AGENTS.md，并在贡献指南和所有受影响的技能中链接到该规则，不重复数字。当这些引用都指向唯一的规则，且如果修改了技能，npm run verify:skills 通过时，任务即完成。
由索引模型根据 Issue 内容生成。
描述
Problem
Nothing in AGENTS.md governs how fast an agent may write to GitHub. The only pacing guidance is in docs/contribution-model.md (closing outside PRs: "a pause of a few seconds between them, in batches"). That is a design document, not loaded every session, and it gives no numbers.
Several flows in this repo make bulk writes, and agents run them:
- the backlog triage (#4875, #4876, #4877): closing PRs with canned comments, boarding and labelling issues
- board audits and recoveries (
board-ops) - Copilot review loops (
pr-flow): many thread replies per round
GitHub's secondary rate limit refuses those writes even while gh api rate_limit still shows primary quota. Most gh writes go through GraphQL, which has its own limits on top. A refused create is also not reliably a no-op: the object can still have been created, so a blind retry duplicates it.
Change
Add this rule to AGENTS.md, worded as given:
Pace bulk GitHub writes: 1 s between requests, never in parallel, under 80 a minute and 500 an hour. The secondary limit refuses while
gh api rate_limitstill shows quota, and mostghwrites are GraphQL. After a refusal, wait outretry-after, or a minute without one; a refused create can still land, so look before retrying.
- The rule belongs in
AGENTS.md(where a rule that must not drop out of context lives), probably under Issue-driven work style or next to Waiting on long-running work, which already sets the 30 s polling floor for remote APIs. docs/contribution-model.md's "Close" step should then link to the rule instead of stating its own pacing, so there is one copy.- Any skill recipe that loops over GitHub writes (
issue-triage,board-ops) should follow it and point to it rather than restating the numbers.
Done when
AGENTS.mdcarries the rule as above.docs/contribution-model.mdand any skill that does bulk writes link to it, with no second copy of the numbers.npm run verify:skillspasses if a skill changed.
- 主要语言
- JavaScript
- 星标
- 91.2k
- 派生
- 11.8k
- 平均合并
- 6 小时 39 分钟
- 30 天内合并 PR
- 111
环境准备
- 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。
modelcontextprotocol/servers 的其他 Issue
-
难度 2/5 1-3 小时 新手友好度 78/100
modelcontextprotocol/servers#5134 ·
维护者通常 1 天内回复
-
sequentialthinking: the thought box's header row is 2 columns narrower than its border可能已有人在做 @nightcityblade 于 2 天前认领。 未关闭bug server-sequentialthinking v2
难度 2/5 1-3 小时 新手友好度 72/100
modelcontextprotocol/servers#5108 · 2 条评论 ·
维护者通常 1 天内回复
-
bug server-git v2
难度 2/5 1-3 小时 新手友好度 72/100
modelcontextprotocol/servers#5107 · 1 条评论 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 75/100
modelcontextprotocol/servers#5071 · 2 条评论 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 85/100
modelcontextprotocol/servers#5059 · 2 条评论 ·
维护者通常 1 天内回复
查看 modelcontextprotocol/servers 的全部 Issue
相似的 Issue
-
难度 2/5 1-3 小时 新手友好度 60/100
quarto-dev/quarto-cli#15042 ·
维护者通常 1 天内回复
-
OpenCode install copies the Pi harness directory (`.pi/`) into the OpenCode config root可能已有人在做 @gaurav0107 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 75/100
维护者通常 3 天内回复
-
难度 2/5 1-3 小时 新手友好度 70/100
NaturalIntelligence/fast-xml-parser#890 · 1 条评论 ·
维护者通常 3 天内回复
-
area: docs enhancement
难度 2/5 1-3 小时 新手友好度 64/100
PyTomography/PyTomography#331 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 68/100
pnpm/get.pnpm.io#68 ·