[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M79A1EV4974FFF01QH0RWNFF":3,"skill-same:01M79A1EV4974FFF01QH0RWNFF":48},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":30,"ownerHtmlUrl":31,"ownerBlogUrl":6,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M79A1EV4974FFF01QH0RWNFF","arbeitsplan-waves",null,"Installs an approved multi-wave design into a project as project-owned files — a pinned interpreter, the plan, a state helper, a path and runner guard, and one agent file per agent type — then launches and resumes it from the primary checkout. Use when a restructuring needs several waves of parallel, disjoint rows, each building on the merged and gated result of the last; when a hand-written multi-wave workflow should become a table; or when a stopped wave run must be resumed. The installed files run without werkstoff. Runs nothing it has not been asked to; a red gate or a pending human gate stops the run, never a retry.","Anselmoo\u002Fwerkstoff","Anselmoo","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F13209783?v=4","01M799YTB3D7RZ4P3B8CSR2X85","Anselm Hahn","01M799YTBRAM6719D9V0NRGB4R",1,"MIT","https:\u002F\u002Fgithub.com\u002FAnselmoo\u002Fwerkstoff\u002Fblob\u002Fmain\u002Fplugins\u002Farbeitsplan\u002Fskills\u002Farbeitsplan-waves\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: arbeitsplan-waves\ndescription: Installs an approved multi-wave design into a project as project-owned files — a pinned interpreter, the plan, a state helper, a path and runner guard, and one agent file per agent type — then launches and resumes it from the primary checkout. Use when a restructuring needs several waves of parallel, disjoint rows, each building on the merged and gated result of the last; when a hand-written multi-wave workflow should become a table; or when a stopped wave run must be resumed. The installed files run without werkstoff. Runs nothing it has not been asked to; a red gate or a pending human gate stops the run, never a retry.\nargument-hint: \"install \u003Cdesign.json> | launch \u003Cname> | resume \u003Cname> | approve \u003Cname> \u003Cgate-id>\"\n---\n\n# Multi-wave runs\n\n`arbeitsplan-run` runs **one** change: N candidates over one scope, and exactly one lands.\nSome restructurings are larger than one change. They need several rows of *different* work\nper wave, each wave built on the merged result of the one before. This skill installs that\nkind of run into the project and drives it. The design comes from `arbeitsplan-design`, and\n`references\u002Fdesign-table-schema.md` is its contract.\n\nEach requirement below comes from a failure observed in a hand-written prototype (#106), and\neach is a check in code rather than a sentence here:\n\n- the run refuses to start from a linked worktree;\n- every row starts with `git merge --ff-only \u003Cwave base>`, and the run halts on a wrong base;\n- a finished wave is skipped before any of its builders is dispatched;\n- merges go into an integration branch, and the target moves only on a green gate;\n- gates run twice, once in the primary checkout and once in a clean worktree, so a finding\n  that only the primary checkout shows is labelled `primary-only`;\n- a swarm row's losers are discarded, never merged;\n- prompts carry ids and branch names only;\n- an **authored** step is written by a shell-less author and verified (syntax, a sample in a\n  scratch worktree, the output schema) before anything runs it, and the guard refuses to run\n  it once its file or its contract has changed.\n\n## Install (once per design)\n\n1. The design must already be approved and written. `compile_spec.py --design --write` prints\n   its sha256. Refuse a design whose hash differs from the approved one.\n\n2. Install:\n\n   ```bash\n   python3 -B \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Finstall_waves.py\" \\\n     --design analysis\u002Farbeitsplan\u002F\u003CrunId>\u002Fdesign.json [--artifact]\n   ```\n\n   Pass `--artifact` when the repository is itself an installable artefact, so the generated\n   tooling is `export-ignore`d.\n\n   A design with **authored steps** gets a skeleton per step under\n   `.claude\u002Fworkflows\u002F\u003Cname>.steps\u002F` (gitignored), in the language its runtime names, plus a\n   `\u003Cname>-author` agent. The first launch writes and verifies them; re-installing never\n   overwrites a step that is no longer a stub.\n\n   The installer resolves the one prerequisite, a Python >= 3.11 for the helper and the guard,\n   by probing `python3`, `python` and `py -3`, or the command given with `--python`. That\n   interpreter is written into every command. The installer refuses to overwrite any file it\n   did not generate.\n\n3. **End the turn. Launch on the next one.** An agent type written to `.claude\u002Fagents\u002F` is not\n   found in the turn that wrote it, not even 30 s later, and resolves from the next turn on\n   (`scripts\u002Fprobe_runtime.py`, probe P4, CLI 2.1.283, ADR 0004). The installer prints\n   `LAUNCH IN A LATER TURN`, and `handoff.py plan` repeats it together with a fresh session's\n   exact start prompt, which also works. Never launch in the turn that ran the install.\n\n## Launch and resume\n\n1. **From the primary checkout.** The wave base is the primary checkout's HEAD, and the\n   merge-gate switches branches there. The interpreter's preflight halts on a linked worktree\n   and on tracked changes; do not try to route around it.\n\n   **Where a builder's worktree starts** was measured by `scripts\u002Fprobe_runtime.py` (P5) on\n   CLI 2.1.283 (ADR 0004). It starts from the project's `worktree.baseRef`: the remote's\n   default branch by default, the primary's HEAD when there is no remote, and the caller's\n   HEAD with `\"head\"`. Each builder then fast-forwards to the wave base. So the base must\n   descend from the remote's default branch: push it, or set `worktree.baseRef` to `\"head\"`\n   in `.claude\u002Fsettings.json`.\n\n2. Read the state and launch:\n\n   ```bash\n   python3 .claude\u002Fworkflows\u002F\u003Cname>_state.py show\n   ```\n\n   Launch the Workflow tool with `scriptPath: .claude\u002Fworkflows\u002F\u003Cname>.js` and\n   `args: {plan: \u003Cthe parsed .claude\u002Fworkflows\u002F\u003Cname>.plan.json>, state: \u003Cthat output>}`.\n   Pass both **as objects**; a JSON string is refused. There is no cost cap by default.\n   Subscription limits stop the run, and relaunching with the saved state is the whole resume.\n\n3. **Read the result, and act on its shape:**\n\n   | result | what to do |\n   |---|---|\n   | `aborted: false` | report the final `integrationSha` and that the target moved; done |\n   | `pending_human_gate: \u003Cid>` | show the user what that gate approves. On their yes, run `\u003Cname>_state.py approve --gate \u003Cid>`, then relaunch |\n   | `abortReason` names a red gate | report `findings` with each one's `source` (`primary-only` means only the primary checkout's untracked or ignored files produced it) and the `kept` worktrees. **Do not retry.** A fix is a new row or a new design |\n   | `PRIMARY CHECKOUT ONLY` | the launch site is wrong; relaunch from the primary checkout |\n   | `WRONG BASE` | a builder could not fast-forward to the wave base (see step 1): push the base, or set `worktree.baseRef` to `\"head\"`, then relaunch |\n   | `SCRIPT CONTRACT` | a declared command exited unexpectedly or printed output that breaks its schema: a defect in the step, not a flake |\n   | `AUTHOR CONTRACT` | an authored step was refused by `verify-step` after its node's `retries`. Report the `problem` and the `step` path. Read the file; the fix is a sharper `purpose`, a better `sample`, or more `retries` in a new design, never a hand edit the hash would refuse anyway |\n\n   `show` reports each authored step as `missing`, `stub`, `authored`, `verified` or `stale`.\n   Only `verified` is skipped on the next launch. To re-check a step by hand, run\n   `\u003Cname>_state.py verify-step --node \u003Cid>`.\n\n   Whatever the result, render where the run stands, from the state it left:\n\n   ```bash\n   python3 .claude\u002Fworkflows\u002F\u003Cname>_state.py show > analysis\u002Farbeitsplan\u002F\u003CrunId>\u002Fshow.json\n   python3 -B \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fbuild_design_html.py\" \\\n     --design .claude\u002Fworkflows\u002F\u003Cname>.plan.json --state analysis\u002Farbeitsplan\u002F\u003CrunId>\u002Fshow.json \\\n     --out analysis\u002Farbeitsplan\u002F\u003CrunId>\u002Fdesign-report.html\n   ```\n\n   Red gates, stale or missing authored steps and pending human gates are marked on the graph\n   where they sit. Quote the printed verdict; it is computed from the state, not from the chat.\n\n   The state helper persists its state as the run goes. Every finished builder is recorded the\n   moment it returns, and every green wave when its gate passes, so a relaunch skips both.\n\n## Rules\n\n- **Never merge by hand to \"help\" a stopped run.** The integration branch and the kept\n  worktrees are the evidence; the next launch or a new design moves forward.\n- **Never widen `offLimits` or edit the generated guard** to get a row through. The guard is a\n  hook because a permission rule did not hold inside agent worktrees.\n- **The installed files are the project's.** A new round of work is a new design and a new\n  install, never an edit to the generated interpreter.\n\n## Output format\n\n```\narbeitsplan-waves rebuild-cli — launched from \u002Fwork\u002Fproj (primary checkout)\n  preflight   head 4e1a9c0, clean, python 3.12.3\n  wave 1      w1-parse ✓  w1-render ✓   gate-1  green (go vet, go test: both trees)   main → 7b2d110\n  wave 2      w2-cli ✓                  gate-2  RED\n                test   primary-only  exit 1   (clean worktree: exit 0)\n              kept: agent\u002Fw2-cli (worktree .claude\u002Fworktrees\u002Fw2-cli)\nSTOPPED at gate-2: main is still 7b2d110. The failure appears only in the primary checkout —\nan untracked or ignored file there. Inspect it; this run does not retry.\nResume after the fix: relaunch with `rebuild-cli_state.py show` — wave 1 will be skipped.\n```\n\n## Resources\n\n- `references\u002Fdesign-table-schema.md` — the design, the wave rules, the rejections.\n- `references\u002Fpatterns.md` — `gated-disjoint-waves`, and why only this form of\n  merging different work is admissible.\n- `scripts\u002Finstall_waves.py`, `scripts\u002Fwaves_state.py`, `scripts\u002Fwaves_guard.py`,\n  `workflows\u002Fwaves.js` — what gets installed, each with its own selftest.\n- `scripts\u002Fbuild_design_html.py` — the plan and its state as a graph.\n","# Multi-wave runs\n\n`arbeitsplan-run` runs **one** change: N candidates over one scope, and exactly one lands.\nSome restructurings are larger than one change. They need several rows of *different* work\nper wave, each wave built on the merged result of the one before. This skill installs that\nkind of run into the project and drives it. The design comes from `arbeitsplan-design`, and\n`references\u002Fdesign-table-schema.md` is its contract.\n\nEach requirement below comes from a failure observed in a hand-written prototype (#106), and\neach is a check in code rather than a sentence here:\n\n- the run refuses to start from a linked worktree;\n- every row starts with `git merge --ff-only \u003Cwave base>`, and the run halts on a wrong base;\n- a finished wave is skipped before any of its builders is dispatched;\n- merges go into an integration branch, and the target moves only on a green gate;\n- gates run twice, once in the primary checkout and once in a clean worktree, so a finding\n  that only the primary checkout shows is labelled `primary-only`;\n- a swarm row's losers are discarded, never merged;\n- prompts carry ids and branch names only;\n- an **authored** step is written by a shell-less author and verified (syntax, a sample in a\n  scratch worktree, the output schema) before anything runs it, and the guard refuses to run\n  it once its file or its contract has changed.\n\n## Install (once per design)\n\n1. The design must already be approved and written. `compile_spec.py --design --write` prints\n   its sha256. Refuse a design whose hash differs from the approved one.\n\n2. Install:\n\n   ```bash\n   python3 -B \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Finstall_waves.py\" \\\n     --design analysis\u002Farbeitsplan\u002F\u003CrunId>\u002Fdesign.json [--artifact]\n   ```\n\n   Pass `--artifact` when the repository is itself an installable artefact, so the generated\n   tooling is `export-ignore`d.\n\n   A design with **authored steps** gets a skeleton per step under\n   `.claude\u002Fworkflows\u002F\u003Cname>.steps\u002F` (gitignored), in the language its runtime names, plus a\n   `\u003Cname>-author` agent. The first launch writes and verifies them; re-installing never\n   overwrites a step that is no longer a stub.\n\n   The installer resolves the one prerequisite, a Python >= 3.11 for the helper and the guard,\n   by probing `python3`, `python` and `py -3`, or the command given with `--python`. That\n   interpreter is written into every command. The installer refuses to overwrite any file it\n   did not generate.\n\n3. **End the turn. Launch on the next one.** An agent type written to `.claude\u002Fagents\u002F` is not\n   found in the turn that wrote it, not even 30 s later, and resolves from the next turn on\n   (`scripts\u002Fprobe_runtime.py`, probe P4, CLI 2.1.283, ADR 0004). The installer prints\n   `LAUNCH IN A LATER TURN`, and `handoff.py plan` repeats it together with a fresh session's\n   exact start prompt, which also works. Never launch in the turn that ran the install.\n\n## Launch and resume\n\n1. **From the primary checkout.** The wave base is the primary checkout's HEAD, and the\n   merge-gate switches branches there. The interpreter's preflight halts on a linked worktree\n   and on tracked changes; do not try to route around it.\n\n   **Where a builder's worktree starts** was measured by `scripts\u002Fprobe_runtime.py` (P5) on\n   CLI 2.1.283 (ADR 0004). It starts from the project's `worktree.baseRef`: the remote's\n   default branch by default, the primary's HEAD when there is no remote, and the caller's\n   HEAD with `\"head\"`. Each builder then fast-forwards to the wave base. So the base must\n   descend from the remote's default branch: push it, or set `worktree.baseRef` to `\"head\"`\n   in `.claude\u002Fsettings.json`.\n\n2. Read the state and launch:\n\n   ```bash\n   python3 .claude\u002Fworkflows\u002F\u003Cname>_state.py show\n   ```\n\n   Launch the Workflow tool with `scriptPath: .claude\u002Fworkflows\u002F\u003Cname>.js` and\n   `args: {plan: \u003Cthe parsed .claude\u002Fworkflows\u002F\u003Cname>.plan.json>, state: \u003Cthat output>}`.\n   Pass both **as objects**; a JSON string is refused. There is no cost cap by default.\n   Subscription limits stop the run, and relaunching with the saved state is the whole resume.\n\n3. **Read the result, and act on its shape:**\n\n   | result | what to do |\n   |---|---|\n   | `aborted: false` | report the final `integrationSha` and that the target moved; done |\n   | `pending_human_gate: \u003Cid>` | show the user what that gate approves. On their yes, run `\u003Cname>_state.py approve --gate \u003Cid>`, then relaunch |\n   | `abortReason` names a red gate | report `findings` with each one's `source` (`primary-only` means only the primary checkout's untracked or ignored files produced it) and the `kept` worktrees. **Do not retry.** A fix is a new row or a new design |\n   | `PRIMARY CHECKOUT ONLY` | the launch site is wrong; relaunch from the primary checkout |\n   | `WRONG BASE` | a builder could not fast-forward to the wave base (see step 1): push the base, or set `worktree.baseRef` to `\"head\"`, then relaunch |\n   | `SCRIPT CONTRACT` | a declared command exited unexpectedly or printed output that breaks its schema: a defect in the step, not a flake |\n   | `AUTHOR CONTRACT` | an authored step was refused by `verify-step` after its node's `retries`. Report the `problem` and the `step` path. Read the file; the fix is a sharper `purpose`, a better `sample`, or more `retries` in a new design, never a hand edit the hash would refuse anyway |\n\n   `show` reports each authored step as `missing`, `stub`, `authored`, `verified` or `stale`.\n   Only `verified` is skipped on the next launch. To re-check a step by hand, run\n   `\u003Cname>_state.py verify-step --node \u003Cid>`.\n\n   Whatever the result, render where the run stands, from the state it left:\n\n   ```bash\n   python3 .claude\u002Fworkflows\u002F\u003Cname>_state.py show > analysis\u002Farbeitsplan\u002F\u003CrunId>\u002Fshow.json\n   python3 -B \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fbuild_design_html.py\" \\\n     --design .claude\u002Fworkflows\u002F\u003Cname>.plan.json --state analysis\u002Farbeitsplan\u002F\u003CrunId>\u002Fshow.json \\\n     --out analysis\u002Farbeitsplan\u002F\u003CrunId>\u002Fdesign-report.html\n   ```\n\n   Red gates, stale or missing authored steps and pending human gates are marked on the graph\n   where they sit. Quote the printed verdict; it is computed from the state, not from the chat.\n\n   The state helper persists its state as the run goes. Every finished builder is recorded the\n   moment it returns, and every green wave when its gate passes, so a relaunch skips both.\n\n## Rules\n\n- **Never merge by hand to \"help\" a stopped run.** The integration branch and the kept\n  worktrees are the evidence; the next launch or a new design moves forward.\n- **Never widen `offLimits` or edit the generated guard** to get a row through. The guard is a\n  hook because a permission rule did not hold inside agent worktrees.\n- **The installed files are the project's.** A new round of work is a new design and a new\n  install, never an edit to the generated interpreter.\n\n## Output format\n\n```\narbeitsplan-waves rebuild-cli — launched from \u002Fwork\u002Fproj (primary checkout)\n  preflight   head 4e1a9c0, clean, python 3.12.3\n  wave 1      w1-parse ✓  w1-render ✓   gate-1  green (go vet, go test: both trees)   main → 7b2d110\n  wave 2      w2-cli ✓                  gate-2  RED\n                test   primary-only  exit 1   (clean worktree: exit 0)\n              kept: agent\u002Fw2-cli (worktree .claude\u002Fworktrees\u002Fw2-cli)\nSTOPPED at gate-2: main is still 7b2d110. The failure appears only in the primary checkout —\nan untracked or ignored file there. Inspect it; this run does not retry.\nResume after the fix: relaunch with `rebuild-cli_state.py show` — wave 1 will be skipped.\n```\n\n## Resources\n\n- `references\u002Fdesign-table-schema.md` — the design, the wave rules, the rejections.\n- `references\u002Fpatterns.md` — `gated-disjoint-waves`, and why only this form of\n  merging different work is admissible.\n- `scripts\u002Finstall_waves.py`, `scripts\u002Fwaves_state.py`, `scripts\u002Fwaves_guard.py`,\n  `workflows\u002Fwaves.js` — what gets installed, each with its own selftest.\n- `scripts\u002Fbuild_design_html.py` — the plan and its state as a graph.\n",{"name":5,"description":7,"argument-hint":24},"install \u003Cdesign.json> | launch \u003Cname> | resume \u003Cname> | approve \u003Cname> \u003Cgate-id>","main","6158c1e4e112c06af2c7ac01b13b2ed7214c0fcf","plugins\u002Farbeitsplan\u002Fskills\u002Farbeitsplan-waves",0,"https:\u002F\u002Fgithub.com\u002FAnselmoo\u002Fwerkstoff","https:\u002F\u002Fanselmoo.github.io\u002Fwerkstoff\u002F","https:\u002F\u002Fgithub.com\u002FAnselmoo","Likes to Develop","2026-11-02 04:20:59",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add Anselmoo\u002Fwerkstoff --skill arbeitsplan-waves",{"tool":40,"command":41},"bunx","bunx skills add Anselmoo\u002Fwerkstoff --skill arbeitsplan-waves",{"tool":43,"command":44},"pnpm","pnpm dlx skills add Anselmoo\u002Fwerkstoff --skill arbeitsplan-waves","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FAnselmoo\u002Fwerkstoff\u002Fblob\u002Fmain\u002Fplugins\u002Farbeitsplan\u002Fskills\u002Farbeitsplan-waves\u002FSKILL.md",true,[],[49,56,63,68,73,78],{"skillNo":50,"name":51,"nameZh":6,"description":52,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":19},"01M79AJTRBNBX03WWKYJ8ENM45","toy-check","A skill that checks things in the repository and reports what it finds.","https:\u002F\u002Fgithub.com\u002FAnselmoo\u002Fwerkstoff\u002Fblob\u002Fmain\u002Ftest\u002Fworkflows\u002Ffixtures\u002Ftoy-plugin\u002Fskills\u002Ftoy-check\u002FSKILL.md",[],[],{"skillNo":57,"name":58,"nameZh":6,"description":59,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":19},"01M79AJS292F3HTDRXWZVRSN5B","demo-check","Reports which generated files changed since the last generation run. Use when the user asks what drifted. Read-only; never edits a file.","https:\u002F\u002Fgithub.com\u002FAnselmoo\u002Fwerkstoff\u002Fblob\u002Fmain\u002Ftest\u002Fplugins\u002Ffixtures\u002Fnacharbeit-uncompleted-run\u002Fplugins\u002Fdemo\u002Fskills\u002Fdemo-check\u002FSKILL.md",[],[],{"skillNo":64,"name":58,"nameZh":6,"description":59,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":65,"categories":66,"keywords":67,"publishedAt":19},"01M79AJP4TFQXKT1AAK1WC1JYK","https:\u002F\u002Fgithub.com\u002FAnselmoo\u002Fwerkstoff\u002Fblob\u002Fmain\u002Ftest\u002Fplugins\u002Ffixtures\u002Fnacharbeit-stale-lock\u002Fplugins\u002Fdemo\u002Fskills\u002Fdemo-check\u002FSKILL.md",[],[],{"skillNo":69,"name":58,"nameZh":6,"description":59,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":19},"01M79AJM6Y50WP3G2BR4H0M60Z","https:\u002F\u002Fgithub.com\u002FAnselmoo\u002Fwerkstoff\u002Fblob\u002Fmain\u002Ftest\u002Fplugins\u002Ffixtures\u002Fnacharbeit-preflight-mixed\u002Fplugins\u002Fdemo\u002Fskills\u002Fdemo-check\u002FSKILL.md",[],[],{"skillNo":74,"name":58,"nameZh":6,"description":59,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":19},"01M79AJGR6DMT8KCCK071V78XM","https:\u002F\u002Fgithub.com\u002FAnselmoo\u002Fwerkstoff\u002Fblob\u002Fmain\u002Ftest\u002Fplugins\u002Ffixtures\u002Fnacharbeit-hooks-shape\u002Fplugins\u002Fdemo\u002Fskills\u002Fdemo-check\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":19},"01M79AJBJ4CHC4PZVWKERWF3HF","zirkel-verify-assumptions","Checks exactly ONE named assumption against real evidence in at most 3 steps and reports one outcome: confidence raised (>=90), reading changed, or still unresolved. Use to resolve a single load-bearing uncertainty — typically a blocking entry surfaced by zirkel-clarify-scope: \"verify this one assumption\", \"is it actually true that X\", \"check this before we rely on it\". For multiple uncertainties, invoke once per uncertainty — the step budget is never shared.","https:\u002F\u002Fgithub.com\u002FAnselmoo\u002Fwerkstoff\u002Fblob\u002Fmain\u002Fplugins\u002Fzirkel\u002Fskills\u002Fzirkel-verify-assumptions\u002FSKILL.md",[],[]]