[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7F3YHSMQM3ZX1P898B4T83Z":3,"skill-same:01M7F3YHSMQM3ZX1P898B4T83Z":45},{"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":24,"commitSha":25,"skillDirectory":26,"forksCount":14,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":29,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M7F3YHSMQM3ZX1P898B4T83Z","work",null,"Build an approved plan from docs\u002Fplans\u002F on its own short-lived branch, or a change the user said to make without a plan. Use when a plan reads state: approved and the user wants it built, when resuming such a build, or when the user has waived the plan for a named change. Work step of the methodical loop. Not when no approved plan exists and the user has not waived one (run the plan step), and not for reviewing finished work.","BigLeverVentures\u002Fmethodical-gated-engineering","BigLeverVentures","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F196365016?v=4","01M7F3V567G0YQ3QWY39AHD9BN","Big Lever Ventures","01M7F3V56S1VHYERWFCQM018V7",0,"MIT","https:\u002F\u002Fgithub.com\u002FBigLeverVentures\u002Fmethodical-gated-engineering\u002Fblob\u002Fmain\u002Fplugins\u002Fmethodical-codex\u002Fskills\u002Fwork\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: work\ndescription: \"Build an approved plan from docs\u002Fplans\u002F on its own short-lived branch, or a change the user said to make without a plan. Use when a plan reads state: approved and the user wants it built, when resuming such a build, or when the user has waived the plan for a named change. Work step of the methodical loop. Not when no approved plan exists and the user has not waived one (run the plan step), and not for reviewing finished work.\"\n---\n\n# Work\n\nSay `Step: work` before anything else, and name the plan file being built (or \"plan waived\").\n\nBuild from a plan in `docs\u002Fplans\u002F` whose front matter reads `state: approved`, or from the user's\nown words in chat to build this change without one. Otherwise do not build: no plan, run the plan\nstep in the same turn; not yet reviewed, deepen if its triggers hold, then run plan-review; awaiting approval, ask \"Approve this plan?\"\nwith the native structured-question tool, as its own final call after the open-question answers are written into the plan, and stop. None of these is offered as a choice. Why: approval is the user's ruling that this plan gets built, and a waiver is the same ruling\ngiven without a plan; only the user gives either.\n\n**Plan waived.** No plan file; go straight to work. Branch as usual (branch hygiene is not waived unless the user says\nso); the user's request, verbatim, is the approach and the acceptance test.\n\n## The escalation line\n\nThe first `Step: work` of a unit says the level from the approved plan's `escalation:` line, with\nthe deepen score when the plan records it: `escalation: lean (deepen 3 of 6)`. It reads three cases\nas `none`, each said as such: `escalation: none (plan waived)`, which gives no recommendation;\n`escalation: none (no line)`, for a plan written before the line existed; and `escalation: none\n(unknown value 'x')`.\n\nAt `lean` or `full` the same line gives one nonblocking recommendation, the first time in a session only:\n\"Session for this build: use the resolved host coordinator model and effort, if not already.\"\nRead the leader setting from the native routing table; use the host profile for worker and reviewer seats.\nModel and effort are stated apart. The user's own settings win: never repeat it, block on it or force a change.\nKeep it nonblocking; a later unit in the same session says nothing about it. It replaces the startup Offer in that line.\nThe parent coordinates the build; the host profile routes each worker for its assigned task.\nWhy: the session's model and effort stay the user's.\n\n## Branch hygiene\n\n`main` is the single trunk. Each unit gets one short-lived branch off `main`, named for the actual\nwork (not an aspirational end-state), and merged back and deleted once its work review clears, its\ntests are green and the user says go. Never commit to `main` directly while a branch for the same\nwork is open; do not base a branch on another unmerged branch, or keep a long-lived integration\nbranch. Why: a branch that outlives its unit drifts from `main`, and the merge then hides what\nchanged.\n\n## Coordinate named workers\n\nFor substantial approved implementation, the parent owns the plan, decisions, worker\ndispatch and recovery, integration of accepted patches, and verification. Assign\nproject code, including material review fixes or new integration code, to named workers.\nUse stable task IDs from the plan as the worker names. Keep the main candidate fixed\nwhile workers produce their own patches in isolated staging directories. Verify that\neach patch matches its approved paths and inputs before accepting it. Recheck its\naccepted digest immediately before integration.\n\nList dependencies, write scopes, acceptance checks and task complexity before dispatch.\nUse the host profile for routine and demanding worker assignments. On a supported host,\nbind the worker guard to the coordinator and launch the full safe ready group before\nwaiting for any one result. Refill newly free slots, including while another task is\nstalled. Record the exact dependency, write overlap, active recovery, or host capacity\nthat requires serial execution. Use one worker if a single cohesive scope has no safe\nsplit. Keep independent reviews, material-fix re-reviews and any required backstop.\n\n## How\n\n1. **Follow the plan's approach in order.** If the plan turns out to be wrong, make the smallest\n   change faithful to its goal, record it as a deviation in the plan, and carry on; ask with the native structured-question tool only if the deviation\n   would change what the user ruled. Why: small deviations are normal; silent ones are\n   how a reviewed plan stops describing the code.\n2. **Check the North Star after each act that changes behaviour**: a census seat (the census row's\n   model and effort) compares the diff, the plan's rulings and its North Star statement, and answers \"on course\" or \"drifted at \u003Cline>,\n   because ...\", quoting the statement line it measured against. Docs-only fixes skip it. Why:\n   drift is cheap to correct one act in, and expensive after the review.\n3. **Test, then commit.** Before every commit run all of the project's tests, unless you have read\n   `${PLUGIN_ROOT}\u002Freference\u002Fnarrow-checks.md` and followed it for this commit, printing each path beside the\n   `narrow-check:` line it matched; a run without that printout is a full run. Then commit, then `git push -u origin \u003Cbranch>`:\n   the unit's branch, never `main`. A red test is fixed, not skipped. Red CI after the push: read the log, decide flake or\n   real, fix, push again. Why: each\n   commit is a point a later session or reviewer trusts as working; a skipped red test hides a break\n   inside history that looks finished; a pushed branch survives the machine, so a unit whose review\n   the user skips is still pushed.\n\n## Owed verification\n\nIf an acceptance test needs real hardware, an environment this machine does not have (a phone, a\ndevice, a cloud account), a live session or a time not yet come, do not leave a note in whatever file is open. Add an entry to the repo's\n`docs\u002Fdevice-verification-queue.md`, creating it from\n`${PLUGIN_ROOT}\u002Ftemplates\u002Fdevice-verification-queue.md` if it is missing: the exact protocol\nin order, what a pass proves, what a fail proves, and the cost. The unit is then **done,\nowed**: it ships, but it is not done until the entry clears. Drain the queue at the next\nnatural moment the hardware is to hand, and before any archive or store build; a fail turns the\nentry into a todo. Why: a note in a random file is never found again, and a cheap exact protocol\nis one that actually gets run.\n\n## When it hands on\n\nWhen the plan's approach is done and the checks that ran pass, say `Step: work-review` and run\n`\u002Fmethodical:work-review` in the same turn. A test that can only run later (hardware, a live\nsession, a time not yet come) is recorded as owed (above) and does not hold the review. Never skip the work review on your own. Why: the author is the one person who cannot\nsee their own gaps, however careful the work. Never merge to `main` or delete a branch without the\nuser's go — compound asks for it once the write-up is on the branch, or a standing merge grant\n(compound skill) gives it, and any merge first runs compound's full run of all of the project's tests. Why: `main` is the\nuser's trunk, and merging is their ruling point.\n","# Work\n\nSay `Step: work` before anything else, and name the plan file being built (or \"plan waived\").\n\nBuild from a plan in `docs\u002Fplans\u002F` whose front matter reads `state: approved`, or from the user's\nown words in chat to build this change without one. Otherwise do not build: no plan, run the plan\nstep in the same turn; not yet reviewed, deepen if its triggers hold, then run plan-review; awaiting approval, ask \"Approve this plan?\"\nwith the native structured-question tool, as its own final call after the open-question answers are written into the plan, and stop. None of these is offered as a choice. Why: approval is the user's ruling that this plan gets built, and a waiver is the same ruling\ngiven without a plan; only the user gives either.\n\n**Plan waived.** No plan file; go straight to work. Branch as usual (branch hygiene is not waived unless the user says\nso); the user's request, verbatim, is the approach and the acceptance test.\n\n## The escalation line\n\nThe first `Step: work` of a unit says the level from the approved plan's `escalation:` line, with\nthe deepen score when the plan records it: `escalation: lean (deepen 3 of 6)`. It reads three cases\nas `none`, each said as such: `escalation: none (plan waived)`, which gives no recommendation;\n`escalation: none (no line)`, for a plan written before the line existed; and `escalation: none\n(unknown value 'x')`.\n\nAt `lean` or `full` the same line gives one nonblocking recommendation, the first time in a session only:\n\"Session for this build: use the resolved host coordinator model and effort, if not already.\"\nRead the leader setting from the native routing table; use the host profile for worker and reviewer seats.\nModel and effort are stated apart. The user's own settings win: never repeat it, block on it or force a change.\nKeep it nonblocking; a later unit in the same session says nothing about it. It replaces the startup Offer in that line.\nThe parent coordinates the build; the host profile routes each worker for its assigned task.\nWhy: the session's model and effort stay the user's.\n\n## Branch hygiene\n\n`main` is the single trunk. Each unit gets one short-lived branch off `main`, named for the actual\nwork (not an aspirational end-state), and merged back and deleted once its work review clears, its\ntests are green and the user says go. Never commit to `main` directly while a branch for the same\nwork is open; do not base a branch on another unmerged branch, or keep a long-lived integration\nbranch. Why: a branch that outlives its unit drifts from `main`, and the merge then hides what\nchanged.\n\n## Coordinate named workers\n\nFor substantial approved implementation, the parent owns the plan, decisions, worker\ndispatch and recovery, integration of accepted patches, and verification. Assign\nproject code, including material review fixes or new integration code, to named workers.\nUse stable task IDs from the plan as the worker names. Keep the main candidate fixed\nwhile workers produce their own patches in isolated staging directories. Verify that\neach patch matches its approved paths and inputs before accepting it. Recheck its\naccepted digest immediately before integration.\n\nList dependencies, write scopes, acceptance checks and task complexity before dispatch.\nUse the host profile for routine and demanding worker assignments. On a supported host,\nbind the worker guard to the coordinator and launch the full safe ready group before\nwaiting for any one result. Refill newly free slots, including while another task is\nstalled. Record the exact dependency, write overlap, active recovery, or host capacity\nthat requires serial execution. Use one worker if a single cohesive scope has no safe\nsplit. Keep independent reviews, material-fix re-reviews and any required backstop.\n\n## How\n\n1. **Follow the plan's approach in order.** If the plan turns out to be wrong, make the smallest\n   change faithful to its goal, record it as a deviation in the plan, and carry on; ask with the native structured-question tool only if the deviation\n   would change what the user ruled. Why: small deviations are normal; silent ones are\n   how a reviewed plan stops describing the code.\n2. **Check the North Star after each act that changes behaviour**: a census seat (the census row's\n   model and effort) compares the diff, the plan's rulings and its North Star statement, and answers \"on course\" or \"drifted at \u003Cline>,\n   because ...\", quoting the statement line it measured against. Docs-only fixes skip it. Why:\n   drift is cheap to correct one act in, and expensive after the review.\n3. **Test, then commit.** Before every commit run all of the project's tests, unless you have read\n   `${PLUGIN_ROOT}\u002Freference\u002Fnarrow-checks.md` and followed it for this commit, printing each path beside the\n   `narrow-check:` line it matched; a run without that printout is a full run. Then commit, then `git push -u origin \u003Cbranch>`:\n   the unit's branch, never `main`. A red test is fixed, not skipped. Red CI after the push: read the log, decide flake or\n   real, fix, push again. Why: each\n   commit is a point a later session or reviewer trusts as working; a skipped red test hides a break\n   inside history that looks finished; a pushed branch survives the machine, so a unit whose review\n   the user skips is still pushed.\n\n## Owed verification\n\nIf an acceptance test needs real hardware, an environment this machine does not have (a phone, a\ndevice, a cloud account), a live session or a time not yet come, do not leave a note in whatever file is open. Add an entry to the repo's\n`docs\u002Fdevice-verification-queue.md`, creating it from\n`${PLUGIN_ROOT}\u002Ftemplates\u002Fdevice-verification-queue.md` if it is missing: the exact protocol\nin order, what a pass proves, what a fail proves, and the cost. The unit is then **done,\nowed**: it ships, but it is not done until the entry clears. Drain the queue at the next\nnatural moment the hardware is to hand, and before any archive or store build; a fail turns the\nentry into a todo. Why: a note in a random file is never found again, and a cheap exact protocol\nis one that actually gets run.\n\n## When it hands on\n\nWhen the plan's approach is done and the checks that ran pass, say `Step: work-review` and run\n`\u002Fmethodical:work-review` in the same turn. A test that can only run later (hardware, a live\nsession, a time not yet come) is recorded as owed (above) and does not hold the review. Never skip the work review on your own. Why: the author is the one person who cannot\nsee their own gaps, however careful the work. Never merge to `main` or delete a branch without the\nuser's go — compound asks for it once the write-up is on the branch, or a standing merge grant\n(compound skill) gives it, and any merge first runs compound's full run of all of the project's tests. Why: `main` is the\nuser's trunk, and merging is their ruling point.\n",{"name":5,"description":7},"main","67bcf3c285965a9714d303252d8023a960a82511","plugins\u002Fmethodical-codex\u002Fskills\u002Fwork","https:\u002F\u002Fgithub.com\u002FBigLeverVentures\u002Fmethodical-gated-engineering","https:\u002F\u002Fgithub.com\u002FBigLeverVentures","www.bigleverventures.com","2026-11-15 10:50:39",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add BigLeverVentures\u002Fmethodical-gated-engineering --skill work",{"tool":37,"command":38},"bunx","bunx skills add BigLeverVentures\u002Fmethodical-gated-engineering --skill work",{"tool":40,"command":41},"pnpm","pnpm dlx skills add BigLeverVentures\u002Fmethodical-gated-engineering --skill work","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FBigLeverVentures\u002Fmethodical-gated-engineering\u002Fblob\u002Fmain\u002Fplugins\u002Fmethodical-codex\u002Fskills\u002Fwork\u002FSKILL.md",true,[],[46,52,58,63,70,77],{"skillNo":47,"name":48,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":49,"categories":50,"keywords":51,"publishedAt":19},"01M7F3ZT8YMJ3EZ6RDQ20X5SVF","phrase-outside","https:\u002F\u002Fgithub.com\u002FBigLeverVentures\u002Fmethodical-gated-engineering\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fhand-off\u002Fphrase-outside\u002FSKILL.md",[],[],{"skillNo":53,"name":54,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":55,"categories":56,"keywords":57,"publishedAt":19},"01M7F3ZR6P6W75DY7QVP7TP9NT","named-only","https:\u002F\u002Fgithub.com\u002FBigLeverVentures\u002Fmethodical-gated-engineering\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fhand-off\u002Fnamed-only\u002FSKILL.md",[],[],{"skillNo":59,"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":60,"categories":61,"keywords":62,"publishedAt":19},"01M7F3ZMHBD3K2WMEFC5XS2224","https:\u002F\u002Fgithub.com\u002FBigLeverVentures\u002Fmethodical-gated-engineering\u002Fblob\u002Fmain\u002Fplugins\u002Fmethodical\u002Fskills\u002Fwork\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":19},"01M7F3ZFVSDD9CMC6X63S8TWK5","work-review","Review a finished build on its branch against its plan, with reviewer seats matched to the code's language, before it merges. Use when the work step's tests pass and the plan's approach is done, or when the user asks for a review of a branch's changes. Work-review step of the methodical loop. Not for reviewing a plan.","https:\u002F\u002Fgithub.com\u002FBigLeverVentures\u002Fmethodical-gated-engineering\u002Fblob\u002Fmain\u002Fplugins\u002Fmethodical\u002Fskills\u002Fwork-review\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M7F3ZC9EXRTW3D90JRAV32ED","plan","Write the plan file in docs\u002Fplans\u002F that code changes get before building, unless the user has said to build without one. Use when the user asks to build, fix, edit or change code and docs\u002Fplans\u002F has no approved plan for it, or when a plan must be revised after review. Plan step of the methodical loop. Not for questions or reading code, and not when the user has waived the plan for this change (go to work).","https:\u002F\u002Fgithub.com\u002FBigLeverVentures\u002Fmethodical-gated-engineering\u002Fblob\u002Fmain\u002Fplugins\u002Fmethodical\u002Fskills\u002Fplan\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M7F3Z8YMRZJDESVP61JQHM27","plan-review","Review a plan in docs\u002Fplans\u002F with reviewer seats matched to the project's language, before the user approves it. Use when a plan has just been written or deepened, when a plan was revised after review, or when the user asks for a plan review. Plan-review step of the methodical loop. Not for reviewing built code.","https:\u002F\u002Fgithub.com\u002FBigLeverVentures\u002Fmethodical-gated-engineering\u002Fblob\u002Fmain\u002Fplugins\u002Fmethodical\u002Fskills\u002Fplan-review\u002FSKILL.md",[],[]]