Skip to main content
在 Manus 中运行任何 Skill
一键导入
bildzeitung
GitHub 创作者资料

bildzeitung

按仓库查看 2 个 GitHub 仓库中的 10 个已收集 skills。

已收集 skills
10
仓库
2
更新
2026-07-11
仓库浏览

仓库与代表性 skills

release
软件开发工程师

Propose the next SemVer version from commit history since the latest vX.Y.Z tag, compile the release notes from the resolved bd epics/tickets in that window (itemized + categorized, delivered as the annotated tag body that CI publishes), get both confirmed (or take an explicit version override), and drive scripts/release.sh to cut the release. Thin wrapper — no build logic of its own; scripts/release.sh (lode-0ru.2) owns the actual gate + tag + push. Parses conventional-commit prefixes (feat -> minor, fix -> patch, `!`/BREAKING CHANGE -> major, but pre-1.0 a breaking change bumps MINOR per docs/release.md) and defaults to a PATCH proposal when no commit carries a recognized prefix. Examples — "/release", "/release patch", "/release minor", "/release major", "/release 0.2.0", "cut a release", "what's the next version".

2026-07-11
code
软件开发工程师

Build one or more lode tasks as PRODUCERS in two phases — dispatch the `coding` subagent (Sonnet) to claim a bd issue, build in an isolated worktree, pass the quality gates, push its branch to origin, and hand off at ready-for-code-review; then dispatch the `code-reviewer` subagent (Opus) to fetch that branch and check it out into its own launch worktree, run the technical review (/code-review + /simplify), re-gate, and swap the ticket to ready-for-land. Producers never merge/close/push trunk; a separate `/land` lander does. Every invocation also sweeps for `needs-rebase` tickets first (branches /land kicked back on a conflict) and dispatches a `coding` producer to rebase, re-gate, force-push, and swap each straight back to ready-for-land — self-heals a clean rebase or a mechanical (independent, non-overlapping) conflict on its own; a conflict where the two sides genuinely disagree still needs a human. `/code <id>` (or `/code --single`) is one producer; bare `/code` / `/code --all-ready` / `/code <id> <id> …`

2026-07-10
debate
项目管理专家

Stress-test a work plan, ticket tree, bug-fix approach, or proposed design change before it gets built. Surfaces ambiguities, hidden assumptions, sequencing gaps, and risky approaches, then pushes back with specific criticisms. Invoke when you want hard pushback on a plan rather than agreement. Examples — "debate this plan", "poke holes in this approach", "stress-test these tickets before I build them".

2026-07-09
sweep
其他计算机职业

The third `/loop` leg — a SURFACE-ONLY human-decision surfacer. Scans bd for work that has stopped waiting on a human and nothing else consumes (`land-escalated` branches, `human`-labeled decision tickets, epics ready for a human close-decision), dedups against a durable cross-machine digest issue, and surfaces new items. Writes no `trunk`, makes no decisions, dispatches no builders/landers/auditors. Run self-paced as `/loop 30m /sweep`. Examples — "/sweep", "/loop 30m /sweep", "what needs a human decision right now?", "sweep the human-decision queue".

2026-07-09
epic-audit
项目管理专家

Closing-side completion gate for an epic — the mirror of `debate`. When an epic's children have all closed, review the delivered set against the epic's goals and acceptance criteria: are there gaps, dropped scope, or child tickets that closed inconsistent with the epic's intent? Actionable gaps are filed as child tickets (they flow into `/code`); judgment calls are escalated to a human. Runs once per completed epic, or as a `/loop` sweep. Examples — "/epic-audit", "/epic-audit lode-mkc", "audit the completed epics", "/loop 30m /epic-audit".

2026-07-08
land-review
软件开发工程师

Semantically review a built, ready-for-land branch against its ticket before it lands — the build-side twin of `debate`. Judges a finished branch on whether it should land: acceptance met? scope clean (no silent creep)? design + lode invariants honored? approach right? Returns a structured verdict accept | bounce | escalate with findings, with enough detail to open a rebuild ticket or surface a decision. It is `/land`'s first task, run once per ready-for-land branch by the lander (not the builder). Distinct from the producer's technical review (`/code-review` + `simplify` = bugs/cleanup); this is semantic — "should this land". Examples — "land-review this branch against lode-123", "semantic-review the ready-for-land queue", "should this branch land?".

2026-06-19
已展示 2 / 2 个仓库
已展示全部仓库