sogeisetsu · Daily GitHub activitysogeisetsu 的 GitHub 每日动态

October 5, 20262026年10月5日

@sogeisetsu·2026-10-05 00:00 – 24:00 · UTC+8

AI summaryAI 摘要

34 commits across two repos; 3 issues filed on OpenCode v2 compatibility两个仓库共 34 次提交；就 OpenCode v2 兼容性新开 3 个 issue

On 2026-10-05, the user pushed 34 commits across two repositories, with no pull requests, reviews, or releases.

In sogeisetsu/meld-deepresearch, 33 own commits reworked documentation and a small feature: the landing page was corrected per review and synced with v0.3.0 and the three skills, install docs clarified that the two one-liners are alternatives, AGENTS, FILE_TREE, CHANGELOG, and PLAN were synced with the change set, the README pair was restructured with install added and three stale eval files dropped, claude-deep-research-skill was recorded as MIT rather than unlicensed, content_review was updated to state that --clean accepts the readability pass, example-1 round 6 (repair attempt 1) was recorded in eval notes, and test/ was marked retired in .gitignore.

A single feature commit wove counter-evidence into the gate instead of labelling it.

In sogeisetsu/sogeisetsu, 1 own commit fixed the daily report so a 0 commit badge is no longer coerced to 1, unified badge layout, and forced inline code in AI summaries; 12 additional automated commits were created by github-actions[bot].

2026-10-05 当天，用户在两个仓库共提交 34 次，无 PR、无 review、无 release。

在 sogeisetsu/meld-deepresearch 中，33 次本人提交主要围绕文档与一处小功能：按 review 修正 landing page 并与 v0.3.0 及三个 skills 同步，install 文档说明两条 one-liner 是二选一，AGENTS、FILE_TREE、CHANGELOG、PLAN 与本次变更集同步，README 对重排并新增 install、删除三个过期 eval 文件，claude-deep-research-skill 记录为 MIT 而非 unlicensed，content_review 中说明 --clean 接受 readability pass，eval 记录 example-1 round 6（repair attempt 1）细节，.gitignore 中标记 test/ 已退役；另有一次功能提交将 counter-evidence 编织进 gate 而非仅打标签。

在 sogeisetsu/sogeisetsu 中，1 次本人提交修复 daily 报告：Commits 角标 0 不再被兜底成 1，徽章排版统一，AI 摘要强制内联代码；另有 12 次由 github-actions[bot] 自动提交。

issue 方面，用户新开 OpenSenseNova/SenseNova-Skills#180，讨论 OpenCode V2 下 SKILL.md frontmatter 与运行时假设的兼容性；新开 iOfficeAI/OfficeCLI#458，修复 OpenCode v2 集成中的 skills 安装路径并新增 mcp target；并在 iOfficeAI/OfficeCLI#254 上留言，指出 OpenCode 检测与 skill 安装路径指向 ~/.opencode，而 OpenCode 实际使用 ~/.config/opencode。

Commits提交

Pull requestsPR

IssuesIssue

Reviews评审

Commits提交

sogeisetsu/meld-deepresearchOwned自有
33commits次提交
- docs(page): fix the landing page per review; correct the port wording
- docs(install): state that the two install one-liners are alternatives
- docs(pages): sync the landing page with v0.3.0 and the three skills
- docs(index): sync AGENTS, FILE_TREE, CHANGELOG and PLAN with the change set
- feat(gate): weave counter-evidence instead of labelling it
- docs(readme): restructure README pair, add install, drop three stale eval files
- docs(notice): record claude-deep-research-skill as MIT, not unlicensed
- docs(script): state in content_review that --clean accepts the readability pass
- docs(eval): record example-1 round 6 (repair attempt 1) details
- chore: mark test/ as retired in .gitignore
- feat(read_table): salvage docx_to_md, document the test/ retirement verdicts
- docs(skill): consolidate report-template around the reader-facing shape rules
- docs(eval): example 1 round 7 PASS - repair attempt 2 of 2 on the new delivery form
- feat(delivery): report shape per blind feedback - definitions, real TOC, no apparatus, no failure narration
- docs(eval): record example-1 round-5 FAIL on E, the fix, and example-2 round-5
- feat(delivery): cited copy to .work, weave discipline chapters, block apparatus leaks
- docs(eval): log the scale-redesign baseline freeze and T1 round-4 PASS
- docs(zh): sync the published Chinese README with the three-skill layout
- docs(changelog): record the scale-redesign additions and changes
- docs(skill): clarify that citing an observation cannot trip the jargon gate
- docs(eval): example 1 round-4 blind eval - PASS on first attempt
- feat(gates): soft budgets + extension, evidence-rule rewrite, layered CI
- feat(core): slim SKILL.md, dual-file render, readability pass, README rewrite
- feat(skills): add meld-da and meld-search-academic capability skills
- chore(eval): freeze baseline medians from example1 record as baseline-f05bcbc.json
- docs(eval): final frozen-skill verdicts - T1 r3 PASS, T2 r4 PASS on 0958af1
- fix(skill): business-register guidance for concise formal reports (T2 E axis)
- docs(eval): T1 round 2 PASS on current skill (main 86b472d)
- feat(skill): stdlib-only local table reader + protocol 2a + CI selftest
- docs(eval): score the G axis from process artifacts on both examples (A-E,G verdict complete)
- docs(eval): record independent verification (gates, links, CI-equivalence, merge clean)
- eval(sensenova-live): T2 round-2 PASS (structure fix verified); finalize records
- fix(skill): honor a requested report structure and register (T2 axis-E gap)
sogeisetsu/sogeisetsuOwned自有
1commits次提交
- fix(daily): Commits 角标 0 不再被兜底成 1，徽章排版统一，AI 摘要强制内联代码
12 automated commits by @github-actions[bot] · not counted in the total12 次自动提交，由 @github-actions[bot] 发起 · 不计入总提交数

IssuesIssue

Opened开启OpenCode V2 compatibility: SKILL.md frontmatter and runtime assumptions
OpenSenseNova/SenseNova-Skills·#180·17:45 · UTC+8
Commented评论OpenCode detection & skill install path point to ~/.opencode, but OpenCode uses ~/.config/opencode
iOfficeAI/OfficeCLI·#254·17:28 · UTC+8
Opened开启OpenCode v2 integration: fix skills install path + add mcp target
iOfficeAI/OfficeCLI·#458·17:27 · UTC+8