[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7DEANC5WVCMPYXP98ECHJFP":3,"skill-same:01M7DEANC5WVCMPYXP98ECHJFP":51},{"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":14,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":30,"ownerDescription":6,"updatedAt":31,"install":32,"resources":44},"01M7DEANC5WVCMPYXP98ECHJFP","ai-playbook-configure",null,"Review and safely edit all four project model roles through guided or expert typed chat, with local presentation preferences and explicit save previews.","rudymolt\u002Fagentic-engineering-playbook","rudymolt","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F265119197?v=4","01M7DE88XGMPPY2D2WYNMDHET1","Rudy Molt","01M7DE88YJAVJ9Z5DH6NM1TCWY",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Frudymolt\u002Fagentic-engineering-playbook\u002Fblob\u002Fmain\u002Fv0.5\u002Fskills\u002Fai-playbook-configure\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: ai-playbook-configure\ndescription: Review and safely edit all four project model roles through guided or expert typed chat, with local presentation preferences and explicit save previews.\ndisable-model-invocation: true\n---\n\n# \u002Fai-playbook-configure\n\nConfigure edits **Plan, Build, Verify and escalated repair**, plus five\nstage-owned job bindings, including portable custom sources. QA inherits Verify;\nCoordinator is observed display-only. Personal defaults and named presets are\nlocal. Evidence-backed recommendations are read-only; unavailable-model\nreplacements edit only the draft and still require explicit Apply.\nThis skill never launches a model or edits runtime state.\n\n## Procedure\n\n1. **Read and discover.** Identify the project and installed playbook locator.\n   Use model-router's existing discovery procedure without its selection\u002Flaunch\n   steps. Exclude unavailable, unauthenticated, blocked, identity-unverifiable\n   or role-ineligible routes. Supply a current-availability adapter to the\n   helper's `--discovery-command`, not a static discovery file. Include the\n   current chat's observed identity as optional `coordinator` route metadata in\n   that same fresh discovery, not a user-supplied model or a guessed catalogue\n   default. If unavailable, display unknown; do not switch the Coordinator.\n   On each request,\n   genuinely recheck role + model + runner + reasoning through the existing\n   authority surface, echo the request ID and timestamp the observation inside\n   the helper's clock window. Never redate cached results or treat guidance's\n   24-hour interval as access evidence. Keep bounded discovery local, not in\n   shareable settings. Read the helper contract at\n   `{playbook-path}\u002Fv0.5\u002Fscripts\u002Fplaybook-config.md` using the installed playbook\n   locator, not a path relative to the copied skill. The mockup and generated\n   availability claims are not evidence. Supply an explicit user-local\n   `--preferences-dir` outside the project on every conversation request;\n   never put a private destination in shared configuration. Pass only known\n   goal, task, risk and billing facts in `read`'s `context`. Read existing project context\n   instead of asking it again; unknown billing is a valid answer. Run the\n   helper's `read` action. On\n   blocked\u002Frecovery-required, report its corrective action and offer Reload or\n   Not now. Completion criterion: current values, origins and verified routes.\n   Official evidence uses a bounded external local cache, optionally selected\n   with `--evidence-dir` on every helper request. The live helper defaults to a\n   private user cache; never copy it into project settings or shared evidence.\n   Reuse successful guidance, reasoning\u002Fcapability and pricing checks for at\n   most 24 hours only at Configure or existing lane-discovery checkpoints.\n   First use, newly observed model\u002Fversion and typed `Refresh` check immediately.\n   Display claim-specific successful dates, fingerprints and material `changes`.\n   Unchanged content produces no invented notice. A failed check retains its\n   successful date and labels stale\u002Fincomplete; new-model advice needs adequate\n   evidence. No background watcher, unrelated-turn check or package installation.\n2. **Present project defaults.** Ask only the returned `missing_context`\n   questions through typed `Goal \u003Ccontext>`, `Billing api`, `Billing subscription`,\n   `Billing mixed` or `Billing unknown`, and `Guided` \u002F `Expert`. Goal\u002Fbilling\n   stay out of project writes; billing saves only after explicit local approval.\n   Known local presentation\u002Fbilling skips those questions on reopening.\n   Show all four model\u002Frunner\u002Freasoning choices and effective origins,\n   QA inheriting Verify, observed Coordinator authority\u002Fdate or unknown,\n   retained repair constraints, exact destination\n   `.playbook-config.json` and whether adoption migrates legacy preferences.\n   Legacy state stays byte-for-byte intact, becoming inert default history\n   after adoption. Offer typed `Edit`, `Edit Plan`, `Edit Build`, `Edit Verify`,\n   `Edit Repair`, `Apply`, `Explain \u003Crole>`, `Not now`. Expert presentation is\n   compact, never a different proposal or weaker validation. Label every row\n   as a retained edition\u002Fproject starting choice or an explicit user edit,\n   not a recommendation. Display the separate `recommendations` rows for all\n   four roles: exact model, runner, supported reasoning, project\u002Ftask fit,\n   risk and limitations. No eligible recommendation means unknown, not an\n   invented fallback. The helper retrieves official sources at initial live\n   `read`; `Explain` reuses that proposal without retrieval or invocation.\n   Provider guidance, published API rates and comparable local outcomes are\n   separate claims, each with its own source URL, successful check date and\n   uncertainty. An incomplete check's retrieval time is not a successful\n   evidence date. General reasoning guidance never proves account access.\n   Show API currency, token unit and billing route; show estimates only with\n   labelled input\u002Foutput\u002Fretry assumptions. Subscription\u002Fmixed routes show\n   observed usage\u002Flimits or unknown, never API prices as a subscription bill.\n   Do not promise savings or a cheapest verified outcome, run candidates,\n   grant approval, or silently rewrite any default. Keep local evidence and\n   billing out of shared config and tracked transcripts. `Task coding` \u002F\n   `Task analysis` and `Risk ordinary` \u002F `Risk high` correct task advice only.\n   For an unbootstrapped project, hand the complete role draft to the existing\n   bootstrap approval preview. Do not run a second bootstrap or write project\n   settings here before that gate; reopen after bootstrap and review again.\n   Completion criterion: the user sees scope, origins, migration and destination.\n3. **Edit and preview.** Pass replies through the helper's `reply` action with\n   the previous proposal. When its step is `replacement`, show the unavailable\n   saved value and exactly one verified suitable alternative, including role,\n   runner\u002Freasoning, task fit, costs and evidence limits. Offer typed `Accept\n   replacement`, `Choose another model` or `Not now`. Acceptance edits only the\n   affected role and returns to the complete before\u002Fafter preview. Choosing\n   another opens that role's ordinary editor. With no suitable alternative,\n   show the limitation and restore-access\u002FRefresh\u002Fexplicit-editor recovery path;\n   do not offer acceptance. Multiple unavailable roles are handled in role order\n   on subsequent discovery\u002FApply checks, retaining all unrelated drafts.\n   Pass ordinary editing replies with\n   the previous proposal. `Edit` offers numbered roles. `Edit \u003Crole>` displays\n   verified complete alternatives numbered from 1, showing model, runner and\n   reasoning. `Pick model` offers numbered model, then supported runner, then\n   supported reasoning, and complete identity if otherwise ambiguous. Never\n   silently choose among providers or optional identity metadata. A number edits only the draft\n   and returns the complete before\u002Fafter preview; show retained roles and\n   constraints too. `Explain \u003Crole>` identifies that selected route, availability\n   observation and limitations without selecting or launching anything.\n   `Explain` defaults to Build and reports discovery authority\u002Fdate and explicitly\n   unknown suitability\u002Fcomparable cost; no paid benchmark. With no alternatives,\n   offer Reload or Not now instead of inventing a route. Native controls may\n   mirror, never replace, typed replies at every level, with immediate text\n   fallback. `Back` returns to the full proposal without losing other edits;\n   an invalid reply retains the previous draft under `retained_proposal` and\n   offers `Back`. `Edit` and `Edit \u003Crole>` from that blocked result also reuse\n   the retained draft, never implicitly reload and discard other edits.\n   Reload intentionally discards unsaved draft changes.\n   `Guided` \u002F `Expert` show a separate local preference before\u002Fafter preview\n   with its exact destination and `resolved_destination` behind directory\n   aliases. If directory identity changed, show the corrective message and\n   require the new preview before another Apply; retain unrelated role drafts.\n   Only `Apply preference` saves it using the same\n   durability\u002Fcompletion\u002Frecovery protocol as project defaults; no project\n   draft is saved by that action. Return to the complete role proposal afterward.\n   `Not now` saves nothing pending; if presentation was explicitly saved earlier,\n   explain that it remains saved while the project draft is cancelled.\n   Completion criterion: reviewed draft\n   or cancellation, with nothing saved.\n   Skill editing uses `Edit skills` or `Edit skills \u003Cjob>` for alignment,\n   specification, implementation, code review or application QA. Show all\n   `skill_proposal` and `skill_changes` before\u002Fafter rows in the full proposal.\n   Numbered `skill_options` retain the same verified source prefix\u002Fprovenance.\n   `Choose \u003Cnumber>` edits one job; `Choose \u003Cnumber>,\u003Cnumber>` retains supported\n   alignment context → decisions composition order. `Explain skills \u003Cjob>`\n   shows inputs, outputs, effects, owner and eligibility without invocation.\n   Unknown, changed, colliding or incompatible sources require an explicit\n   fallback or block, never warning acknowledgement. Read\n   [`..\u002F..\u002Fscripts\u002Fskill-bindings.md`](..\u002F..\u002Fscripts\u002Fskill-bindings.md) for\n   contracts, current upstream exclusions and already eligible project QA proof.\n   Configuration does not create or maintain a harness.\n   `Presets` shows one Recommended entry plus named user presets. Recommended\n   retains the starting draft; evidence-backed suggestions are displayed\n   separately and do not select a model. No static cost tiers or\n   availability-as-suitability claim. `Load defaults`\n   or `Load preset \u003Cname>` changes only the draft, showing model\u002Fskill origins,\n   before\u002Fafter changes and destinations. `Save defaults` or `Save preset \u003Cname>`\n   previews personal reusable data using the same validated model\u002Fskill schema.\n   `Apply preference` saves only that personal data; other projects and active\n   work remain unchanged. `Billing \u003Cmode>` previews a local billing preference;\n   unknown consumption remains unknown and API prices are not subscription bills.\n   Presentation edits preserve saved billing, defaults and presets. Missing\n   dependencies or unknown fields require explicit review, not silent omission.\n   Custom sources use a canonical project-relative identity or a logical\n   identity resolved through an external machine-local binding\u002Faudit store.\n   Show `(Custom)` unless existing stage evidence verifies a more specific\n   source label. Missing local resolution\u002Fevidence requires explicit recovery,\n   never a source rewrite or automatic fallback. Local JSON assertions are not\n   real independent execution proof; the owning stage authenticates retained\n   qualification at its existing gate. Never generate that proof by invocation.\n4. **Apply the reviewed proposal.** Only after typed `Apply`, repeat genuine\n   authoritative discovery through that adapter and pass the same proposal to the helper. Never\n   silently create and Apply a refreshed proposal. Changed inputs require\n   Reload and another preview\u002FApply. Unavailability returns the replacement\n   conversation with the draft retained. Acceptance requires a refreshed preview\n   and another explicit Apply; never substitute or save the old proposal.\n   Report applied\u002Funchanged\u002Fblocked\u002Frecovery-required\n   exactly with destination and corrective message. `Not now` writes no\n   pending defaults or presets; distinguish earlier explicitly saved presentation\n   and any local read-side discovery refresh. Stop for retained recovery evidence;\n   never remove it to make a save appear successful.\n   When the preview includes a personal change, project `Apply` uses paired\n   transaction journals and content completion for both destinations. Report\n   every destination and the true result; independent saves are not paired\n   success. Exact rollback still reports recovery-required, and concurrent\n   edits\u002Fcaptured bytes must never be overwritten. Keep journals and private\n   evidence local; never put billing, binding-store paths or unreviewed exception\n   text in shareable settings\u002Fresults. The reviewed personal destination remains\n   visible in the local review\u002Frecovery flow. New-project seeding belongs to\n   the bootstrap skill's existing preview and exact seed approval, never to an\n   implicit Configure bootstrap or stage\u002Fmodel launch.\n   Completion criterion: validated save, no-op, cancellation or named blocker,\n   with no model dispatch or build.\n\nReturn `playbook_result` with outcome applied\u002Funchanged\u002Fblocked,\n`next_stage: null`, and ordered `required_actions` from the corrective result.\nExisting lane gates remain selection, approval and launch authority.\n","# \u002Fai-playbook-configure\n\nConfigure edits **Plan, Build, Verify and escalated repair**, plus five\nstage-owned job bindings, including portable custom sources. QA inherits Verify;\nCoordinator is observed display-only. Personal defaults and named presets are\nlocal. Evidence-backed recommendations are read-only; unavailable-model\nreplacements edit only the draft and still require explicit Apply.\nThis skill never launches a model or edits runtime state.\n\n## Procedure\n\n1. **Read and discover.** Identify the project and installed playbook locator.\n   Use model-router's existing discovery procedure without its selection\u002Flaunch\n   steps. Exclude unavailable, unauthenticated, blocked, identity-unverifiable\n   or role-ineligible routes. Supply a current-availability adapter to the\n   helper's `--discovery-command`, not a static discovery file. Include the\n   current chat's observed identity as optional `coordinator` route metadata in\n   that same fresh discovery, not a user-supplied model or a guessed catalogue\n   default. If unavailable, display unknown; do not switch the Coordinator.\n   On each request,\n   genuinely recheck role + model + runner + reasoning through the existing\n   authority surface, echo the request ID and timestamp the observation inside\n   the helper's clock window. Never redate cached results or treat guidance's\n   24-hour interval as access evidence. Keep bounded discovery local, not in\n   shareable settings. Read the helper contract at\n   `{playbook-path}\u002Fv0.5\u002Fscripts\u002Fplaybook-config.md` using the installed playbook\n   locator, not a path relative to the copied skill. The mockup and generated\n   availability claims are not evidence. Supply an explicit user-local\n   `--preferences-dir` outside the project on every conversation request;\n   never put a private destination in shared configuration. Pass only known\n   goal, task, risk and billing facts in `read`'s `context`. Read existing project context\n   instead of asking it again; unknown billing is a valid answer. Run the\n   helper's `read` action. On\n   blocked\u002Frecovery-required, report its corrective action and offer Reload or\n   Not now. Completion criterion: current values, origins and verified routes.\n   Official evidence uses a bounded external local cache, optionally selected\n   with `--evidence-dir` on every helper request. The live helper defaults to a\n   private user cache; never copy it into project settings or shared evidence.\n   Reuse successful guidance, reasoning\u002Fcapability and pricing checks for at\n   most 24 hours only at Configure or existing lane-discovery checkpoints.\n   First use, newly observed model\u002Fversion and typed `Refresh` check immediately.\n   Display claim-specific successful dates, fingerprints and material `changes`.\n   Unchanged content produces no invented notice. A failed check retains its\n   successful date and labels stale\u002Fincomplete; new-model advice needs adequate\n   evidence. No background watcher, unrelated-turn check or package installation.\n2. **Present project defaults.** Ask only the returned `missing_context`\n   questions through typed `Goal \u003Ccontext>`, `Billing api`, `Billing subscription`,\n   `Billing mixed` or `Billing unknown`, and `Guided` \u002F `Expert`. Goal\u002Fbilling\n   stay out of project writes; billing saves only after explicit local approval.\n   Known local presentation\u002Fbilling skips those questions on reopening.\n   Show all four model\u002Frunner\u002Freasoning choices and effective origins,\n   QA inheriting Verify, observed Coordinator authority\u002Fdate or unknown,\n   retained repair constraints, exact destination\n   `.playbook-config.json` and whether adoption migrates legacy preferences.\n   Legacy state stays byte-for-byte intact, becoming inert default history\n   after adoption. Offer typed `Edit`, `Edit Plan`, `Edit Build`, `Edit Verify`,\n   `Edit Repair`, `Apply`, `Explain \u003Crole>`, `Not now`. Expert presentation is\n   compact, never a different proposal or weaker validation. Label every row\n   as a retained edition\u002Fproject starting choice or an explicit user edit,\n   not a recommendation. Display the separate `recommendations` rows for all\n   four roles: exact model, runner, supported reasoning, project\u002Ftask fit,\n   risk and limitations. No eligible recommendation means unknown, not an\n   invented fallback. The helper retrieves official sources at initial live\n   `read`; `Explain` reuses that proposal without retrieval or invocation.\n   Provider guidance, published API rates and comparable local outcomes are\n   separate claims, each with its own source URL, successful check date and\n   uncertainty. An incomplete check's retrieval time is not a successful\n   evidence date. General reasoning guidance never proves account access.\n   Show API currency, token unit and billing route; show estimates only with\n   labelled input\u002Foutput\u002Fretry assumptions. Subscription\u002Fmixed routes show\n   observed usage\u002Flimits or unknown, never API prices as a subscription bill.\n   Do not promise savings or a cheapest verified outcome, run candidates,\n   grant approval, or silently rewrite any default. Keep local evidence and\n   billing out of shared config and tracked transcripts. `Task coding` \u002F\n   `Task analysis` and `Risk ordinary` \u002F `Risk high` correct task advice only.\n   For an unbootstrapped project, hand the complete role draft to the existing\n   bootstrap approval preview. Do not run a second bootstrap or write project\n   settings here before that gate; reopen after bootstrap and review again.\n   Completion criterion: the user sees scope, origins, migration and destination.\n3. **Edit and preview.** Pass replies through the helper's `reply` action with\n   the previous proposal. When its step is `replacement`, show the unavailable\n   saved value and exactly one verified suitable alternative, including role,\n   runner\u002Freasoning, task fit, costs and evidence limits. Offer typed `Accept\n   replacement`, `Choose another model` or `Not now`. Acceptance edits only the\n   affected role and returns to the complete before\u002Fafter preview. Choosing\n   another opens that role's ordinary editor. With no suitable alternative,\n   show the limitation and restore-access\u002FRefresh\u002Fexplicit-editor recovery path;\n   do not offer acceptance. Multiple unavailable roles are handled in role order\n   on subsequent discovery\u002FApply checks, retaining all unrelated drafts.\n   Pass ordinary editing replies with\n   the previous proposal. `Edit` offers numbered roles. `Edit \u003Crole>` displays\n   verified complete alternatives numbered from 1, showing model, runner and\n   reasoning. `Pick model` offers numbered model, then supported runner, then\n   supported reasoning, and complete identity if otherwise ambiguous. Never\n   silently choose among providers or optional identity metadata. A number edits only the draft\n   and returns the complete before\u002Fafter preview; show retained roles and\n   constraints too. `Explain \u003Crole>` identifies that selected route, availability\n   observation and limitations without selecting or launching anything.\n   `Explain` defaults to Build and reports discovery authority\u002Fdate and explicitly\n   unknown suitability\u002Fcomparable cost; no paid benchmark. With no alternatives,\n   offer Reload or Not now instead of inventing a route. Native controls may\n   mirror, never replace, typed replies at every level, with immediate text\n   fallback. `Back` returns to the full proposal without losing other edits;\n   an invalid reply retains the previous draft under `retained_proposal` and\n   offers `Back`. `Edit` and `Edit \u003Crole>` from that blocked result also reuse\n   the retained draft, never implicitly reload and discard other edits.\n   Reload intentionally discards unsaved draft changes.\n   `Guided` \u002F `Expert` show a separate local preference before\u002Fafter preview\n   with its exact destination and `resolved_destination` behind directory\n   aliases. If directory identity changed, show the corrective message and\n   require the new preview before another Apply; retain unrelated role drafts.\n   Only `Apply preference` saves it using the same\n   durability\u002Fcompletion\u002Frecovery protocol as project defaults; no project\n   draft is saved by that action. Return to the complete role proposal afterward.\n   `Not now` saves nothing pending; if presentation was explicitly saved earlier,\n   explain that it remains saved while the project draft is cancelled.\n   Completion criterion: reviewed draft\n   or cancellation, with nothing saved.\n   Skill editing uses `Edit skills` or `Edit skills \u003Cjob>` for alignment,\n   specification, implementation, code review or application QA. Show all\n   `skill_proposal` and `skill_changes` before\u002Fafter rows in the full proposal.\n   Numbered `skill_options` retain the same verified source prefix\u002Fprovenance.\n   `Choose \u003Cnumber>` edits one job; `Choose \u003Cnumber>,\u003Cnumber>` retains supported\n   alignment context → decisions composition order. `Explain skills \u003Cjob>`\n   shows inputs, outputs, effects, owner and eligibility without invocation.\n   Unknown, changed, colliding or incompatible sources require an explicit\n   fallback or block, never warning acknowledgement. Read\n   [`..\u002F..\u002Fscripts\u002Fskill-bindings.md`](..\u002F..\u002Fscripts\u002Fskill-bindings.md) for\n   contracts, current upstream exclusions and already eligible project QA proof.\n   Configuration does not create or maintain a harness.\n   `Presets` shows one Recommended entry plus named user presets. Recommended\n   retains the starting draft; evidence-backed suggestions are displayed\n   separately and do not select a model. No static cost tiers or\n   availability-as-suitability claim. `Load defaults`\n   or `Load preset \u003Cname>` changes only the draft, showing model\u002Fskill origins,\n   before\u002Fafter changes and destinations. `Save defaults` or `Save preset \u003Cname>`\n   previews personal reusable data using the same validated model\u002Fskill schema.\n   `Apply preference` saves only that personal data; other projects and active\n   work remain unchanged. `Billing \u003Cmode>` previews a local billing preference;\n   unknown consumption remains unknown and API prices are not subscription bills.\n   Presentation edits preserve saved billing, defaults and presets. Missing\n   dependencies or unknown fields require explicit review, not silent omission.\n   Custom sources use a canonical project-relative identity or a logical\n   identity resolved through an external machine-local binding\u002Faudit store.\n   Show `(Custom)` unless existing stage evidence verifies a more specific\n   source label. Missing local resolution\u002Fevidence requires explicit recovery,\n   never a source rewrite or automatic fallback. Local JSON assertions are not\n   real independent execution proof; the owning stage authenticates retained\n   qualification at its existing gate. Never generate that proof by invocation.\n4. **Apply the reviewed proposal.** Only after typed `Apply`, repeat genuine\n   authoritative discovery through that adapter and pass the same proposal to the helper. Never\n   silently create and Apply a refreshed proposal. Changed inputs require\n   Reload and another preview\u002FApply. Unavailability returns the replacement\n   conversation with the draft retained. Acceptance requires a refreshed preview\n   and another explicit Apply; never substitute or save the old proposal.\n   Report applied\u002Funchanged\u002Fblocked\u002Frecovery-required\n   exactly with destination and corrective message. `Not now` writes no\n   pending defaults or presets; distinguish earlier explicitly saved presentation\n   and any local read-side discovery refresh. Stop for retained recovery evidence;\n   never remove it to make a save appear successful.\n   When the preview includes a personal change, project `Apply` uses paired\n   transaction journals and content completion for both destinations. Report\n   every destination and the true result; independent saves are not paired\n   success. Exact rollback still reports recovery-required, and concurrent\n   edits\u002Fcaptured bytes must never be overwritten. Keep journals and private\n   evidence local; never put billing, binding-store paths or unreviewed exception\n   text in shareable settings\u002Fresults. The reviewed personal destination remains\n   visible in the local review\u002Frecovery flow. New-project seeding belongs to\n   the bootstrap skill's existing preview and exact seed approval, never to an\n   implicit Configure bootstrap or stage\u002Fmodel launch.\n   Completion criterion: validated save, no-op, cancellation or named blocker,\n   with no model dispatch or build.\n\nReturn `playbook_result` with outcome applied\u002Funchanged\u002Fblocked,\n`next_stage: null`, and ordered `required_actions` from the corrective result.\nExisting lane gates remain selection, approval and launch authority.\n",{"name":5,"description":7,"disable-model-invocation":24},true,"main","0f5edd22fa88b047967d9263c1b5f55b91041b22","v0.5\u002Fskills\u002Fai-playbook-configure","https:\u002F\u002Fgithub.com\u002Frudymolt\u002Fagentic-engineering-playbook","https:\u002F\u002Fgithub.com\u002Frudymolt","rudymolt.github.io","2026-11-14 08:49:32",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":24},[34,37,40],{"tool":35,"command":36},"npx","npx skills add rudymolt\u002Fagentic-engineering-playbook --skill ai-playbook-configure",{"tool":38,"command":39},"bunx","bunx skills add rudymolt\u002Fagentic-engineering-playbook --skill ai-playbook-configure",{"tool":41,"command":42},"pnpm","pnpm dlx skills add rudymolt\u002Fagentic-engineering-playbook --skill ai-playbook-configure","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Frudymolt\u002Fagentic-engineering-playbook\u002Fblob\u002Fmain\u002Fv0.5\u002Fskills\u002Fai-playbook-configure\u002FSKILL.md",[45],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"agents\u002Fopenai.yaml","REFERENCE","参考文档",334,"https:\u002F\u002Fgithub.com\u002Frudymolt\u002Fagentic-engineering-playbook\u002Fblob\u002Fmain\u002Fv0.5\u002Fskills\u002Fai-playbook-configure\u002Fagents\u002Fopenai.yaml",[52,59,66,73,80,87],{"skillNo":53,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":19},"01M7DEBXPTYATNMDDZHNA2SG8M","whats-next","Route the next playbook stage from project state and cadence evidence. Use when the user asks what to do next, is stuck between stages, or another playbook rule needs an ambient recommendation.","https:\u002F\u002Fgithub.com\u002Frudymolt\u002Fagentic-engineering-playbook\u002Fblob\u002Fmain\u002Fv0.5\u002Fskills\u002Fwhats-next\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":19},"01M7DEBT6NX786BXE8N5P6Z1RA","ship-release","Publish or reconcile a repository version tag and GitHub Release. Use only after the user explicitly asks to cut or publish a release, or to fix drift between version files, tags, and GitHub Releases.","https:\u002F\u002Fgithub.com\u002Frudymolt\u002Fagentic-engineering-playbook\u002Fblob\u002Fmain\u002Fv0.5\u002Fskills\u002Fship-release\u002FSKILL.md",[],[],{"skillNo":67,"name":68,"nameZh":6,"description":69,"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},"01M7DEBPJ5MAD55H3563Q76724","model-router","Discover verified model routes and launch the selected Plan, Build, or Verify lane. Use at feature-lane boundaries, when the user types models or openai defaults, or when a saved model route must resume.","https:\u002F\u002Fgithub.com\u002Frudymolt\u002Fagentic-engineering-playbook\u002Fblob\u002Fmain\u002Fv0.5\u002Fskills\u002Fmodel-router\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":19},"01M7DEBJ6PANTA3Y629EWYH09D","matt-retro","Review a coding session with Matt Pocock's environment-improvement retrospective.","https:\u002F\u002Fgithub.com\u002Frudymolt\u002Fagentic-engineering-playbook\u002Fblob\u002Fmain\u002Fv0.5\u002Fskills\u002Fmatt-retro\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":19},"01M7DEBE05TH9T3YBB1WNX0YBX","ai-playbook-why","Explain a current decision's historical motivation from bounded, cited evidence while separating direct facts, inference, contradictions, and unknowns.","https:\u002F\u002Fgithub.com\u002Frudymolt\u002Fagentic-engineering-playbook\u002Fblob\u002Fmain\u002Fv0.5\u002Fskills\u002Fai-playbook-why\u002FSKILL.md",[],[],{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":93,"publishedAt":19},"01M7DEBA6J9CGRZR7Z85Z987DQ","ai-playbook-verification-harness","Discover, propose, generate, and prove a project-owned verification harness without replacing its existing commands or claiming a real-project pilot.","https:\u002F\u002Fgithub.com\u002Frudymolt\u002Fagentic-engineering-playbook\u002Fblob\u002Fmain\u002Fv0.5\u002Fskills\u002Fai-playbook-verification-harness\u002FSKILL.md",[],[]]