[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M74ND9DKNF207956FJHTXBXB":3,"skill-same:01M74ND9DKNF207956FJHTXBXB":77},{"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":27,"commitSha":28,"skillDirectory":29,"forksCount":30,"repositoryHtmlUrl":31,"repositoryHomepageUrl":32,"ownerHtmlUrl":33,"ownerBlogUrl":34,"ownerDescription":35,"updatedAt":36,"install":37,"resources":50},"01M74ND9DKNF207956FJHTXBXB","idd",null,"Impact-driven development protocol: run a code change through the ospec CLI, which derives its obligations and decides when it is done. Trigger: a code change, unless SDD is requested or mode: sdd.","Snakeblack\u002Fospec-workflow","Snakeblack","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F49766621?v=4","01M74NA5A4XHJYFRMG9JHM2R4K","Michael Snake","01M74NA5ASKTH7DG1P6DKC8WTX",3,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fidd\u002FSKILL.md",[],[],"2026-11-11 15:11:14",false,"---\nname: idd\ndescription: \"Impact-driven development protocol: run a code change through the ospec CLI, which derives its obligations and decides when it is done. Trigger: a code change, unless SDD is requested or mode: sdd.\"\nlicense: Apache-2.0\nmetadata:\n  author: manuel-retamozo-garcia\n  version: \"1.0\"\n---\n\n## When to Use\n\nLoad this skill once, when the router sends a code change here (IDD is the\ndefault). It is the whole protocol — no phases, proposal or specs first.\nQuestions, explanations and read-only work need no change and do not use it.\n\nOutside a git work tree `ospec` opens no change (`not-a-git-repo`,\n`git-not-found`): set up the repository first, as `references\u002Frepository.md`\nsays. It also covers a change across several services or repositories.\n\n## The CLI decides\n\n`ospec` below means `{{ospec-cli}}`, run from the project root. Add\n`--json` to read its answer; every change command takes `--change \u003Cid>`. The\nCLI owns `idd\u002F\u003Cchange>\u002Fstate.yaml`: never write that file, and never move or\ndelete anything under `idd\u002F` by hand.\n\n- Every command that writes a change prints its next step; `ospec next\n  --change \u003Cid>` gives it in full when you resume. Its `next_step.how` tells\n  how to proceed: run the command that records the evidence or, for\n  `configure-checks`, propose a check command as `references\u002Fchecks.md` says;\n  without one no IDD change can close. Follow it.\n- An obligation is satisfied only by evidence the CLI records from a run it\n  observes (`ospec run`, `ospec check`, `ospec review`). Saying that tests\n  pass satisfies nothing.\n- The change is done when `ospec close` succeeds, and only then. Never report\n  it done, verified or archived before that.\n\n## Rules\n\n1. Read the code the request touches, then list its open facts: each behavior\n   that neither the request nor the code settles and that a careful engineer\n   would ask about. Look for defaults, invalid or edge input, error versus\n   silent handling, normalization (case, rounding, order), who may do it, and\n   whether existing behavior, output or a published API may change. Do not\n   decide them yourself.\n2. Open the change: pick a short kebab-case id from the request and record\n   the intent with `ospec record intent --change \u003Cid> --kind \u003Cbug|feature|refactor|docs> --summary \"\u003Cwhat>\" --acceptance \"\u003Cobservable result>\"`\n   plus the open facts, one `--open-fact \"\u003Cquestion>\"` each, or\n   `--no-open-facts --basis \"\u003Cwhat settles every behavior>\"` when there are\n   none. A `feature` covers the categories of the template `ospec next`\n   prints, as `references\u002Fopen-facts.md` says. Assumptions: `ospec record assumption --text`\n   (close lists them). With cross-session memory tools, recall prior\n   decisions with the `memory_query` it returns, as\n   `references\u002Fmemory-loop.md` says. Open `open-facts`: then ask all at once\n   before editing anything.\n3. If the request is materially ambiguous (no acceptance can be stated), record\n   `ospec record intent --change \u003Cid> --ambiguous --request \"\u003Crequest>\"`, ask\n   the user the question `next` returns, and resolve it with the `record\n   intent` that `next` prints, without `--ambiguous`, filling in their words\n   and the open-facts declaration of rule 2.\n4. Declare the plan before editing, on `record intent` or later with\n   `ospec signals --change \u003Cid>`: `--path \u003Cfile>` once per file you\n   expect to touch; add `--work-units \u003Cn>` when the work\n   splits into several units, `--decision \"\u003Ctext>\"` per design decision,\n   and `--operation \u003Cop>` for a destructive or irreversible operation. With\n   `docs\u002Farchitecture\u002F`, declare the ADRs and quality scenarios the change\n   bears on (`--adr \u003Cid>`, `--quality \u003Cid>`) as `references\u002Farchitecture.md`\n   says. If the plan overstated a signal, retract\n   it with `ospec record retract --change \u003Cid> --signal \u003Cid> --reason \"\u003Cwhy>\"`;\n   the CLI refuses while the diff confirms it. When its `how` offers the\n   user's answer, ask the user; never answer for them.\n5. Loop: do what the `next_step` of the last command says, repeat.\n   `resolve-gate` comes first while any gate is open: ask its\n   `pending_decision` (see Gates) and build nothing it decides until the user\n   answers. Work the obligations in the order `next` gives them:\n   - `repro-test`: write the reproduction test first and run it with\n     `ospec run --obligation repro-test --command \"\u003Ctest>\"` while it fails;\n     fix the code; run the same command again so it passes.\n   - `tdd-red-green`: the same red-then-green pair, per unit (`--unit \u003Cname>`).\n   - `contract-spec-and-test`: update the contract document and its test.\n   - `migration-compat-and-test`: run the migration test with\n     `--plan \"\u003Ccompatibility or rollback plan>\"`.\n   - `adr-impact-declaration`: `ospec record impact` for each declared ADR or\n     scenario, and their fitness checks, as `references\u002Farchitecture.md` says.\n   - `trust-review`: `ospec review start`, hand the whole returned request to\n     the read-only `review-trust` reviewer, and pass its JSON unchanged to\n     `ospec review record --result @\u003Cfile>`. Fix what it blocks on, run\n     `ospec review correct`, hand its request to the read-only\n     `review-correction` validator and pass its result to\n     `ospec review validate --result @\u003Cfile>`.\n   - `checks-pass`: `ospec check` on the current tree, or a run whose command\n     is the only check; a `tests 0` summary is not evidence. `--close`, when\n     its command has it, closes the change once ready.\n   - `living-doc`: the CLI writes the Plan and Decisions of `change.md` from\n     `--path` and `--decision`; add only what they miss. A development\n     decision stays there; it is never an ADR.\n\n   For the `suggestions`, `references` and `warnings` of `next`, read\n   `references\u002Farchitecture.md`; tell the user each warning.\n6. `ospec check` recomputes the signals from the real diff. New obligations can\n   appear: work them like the rest. One leaves `pending` only through evidence,\n   or through `ospec record withdraw --reason` when no active signal derives it.\n7. Edit text files keeping their encoding and line endings: never rewrite a\n   file through a tool or script that changes them (for example Python\n   `open()` without `encoding=` and `newline=`).\n8. When `next_step.action` is `close`, run `ospec close --change \u003Cid>`. A\n   refusal names what is still pending: settle it and close again.\n\n## Build it simply\n\nMake the smallest change that meets the acceptance. Reuse what the code\nalready has and follow its conventions. Add a layer, abstraction, dependency\nor configuration switch only for a need present in this change, and record\nthe choice with `--decision \"\u003Ctext>\"` when it is not obvious. Leave unrelated code as\nit is: report a defect outside the change instead of fixing it in passing.\n\n## Tests by risk\n\nTest what would cost something if it broke: business rules, validation,\nnon-trivial transformations and mappers, algorithms, contracts, and their\nerror, edge and regression paths. Code with nothing to get wrong (plain\naccessors, wiring without logic, declarations) needs no test of its own, and\nno coverage percentage replaces this choice.\n\n- A test must fail when the behavior is wrong: assert observable results, not\n  the calls the code makes, and check that the red run fails for the reason\n  you expect.\n- Unit tests are deterministic and offline: replace databases, external APIs\n  and infrastructure with test doubles at the boundary, and inject time and\n  randomness. An integration test, kept apart, proves the wiring.\n- A refactor writes no new test: run `ospec check` before the first edit, so\n  a recorded run shows the behavior it preserves was tested.\n\n## Gates\n\nStop for the user at the gates: end your turn with the questions, not a\nquestion tool. A `user` answer given before a turn ends is kept\n`user-unverified`.\nA skipped, empty or default answer is not the user's: end your turn with the\nquestions and wait.\n\n- `ambiguous-intent`: rule 3.\n- `open-facts`: rule 2; ask all of `pending_decision.questions` at once.\n- `irreversible-operation`: `--operation` in rule 4 or a destructive\n  statement in the diff opens it. Before any other destructive or\n  irreversible step (deleting data, rewriting published history), open it\n  yourself: `ospec record gate --change \u003Cid> --gate irreversible-operation --open --reason \"\u003Coperation>\"`.\n- `adr-amend-or-contradict`: an impact `amends`, `supersedes` or\n  `contradicts` opens it. Declare every one before asking, since the gate is\n  answered once. For a departure no declared reference covers, open it with\n  `ospec record gate --change \u003Cid> --gate adr-amend-or-contradict --open --reason \"\u003CADR>: \u003Chow the change departs from it>\"`.\n  A change that conforms opens nothing.\n- `review-limit`: `ospec check` opens it when the trust reviews run out.\n- `checks-changed`: `ospec check` opens it, running nothing, when the checks\n  or the ceremony settings (`strict_tdd`, `impact`, `contracts`) differ from\n  the change's base. The user's approval of exactly those values answers it.\n\nResolve each one with `ospec record gate --change \u003Cid> --gate \u003Cid> --resolve --answer \"\u003Ctheir words>\" --source user`,\nor `user-unverified` for a relayed answer. The CLI cannot verify who answered:\nread `references\u002Fapprovals.md` first. Only an explicit answer from the user\nresolves a gate. Outside these gates, follow the code and the evidence\nwithout asking.\n\n## Delivery\n\n`ospec close` archives the change under `idd\u002Farchive\u002F`, adds its entry to the\nchangelog `delivery.changelog` declares, and leaves the rest of the tree as it\nis. Branches, commits, pull requests and releases are the user's decision;\npropose the `team.delivery_strategy` of `next`, and the `delivery` it returns\nwhen the work starts and at close: each of its options, why it fits\n(`reasons`) and the rules to follow (`how`). With a\n`*-undeclared` suggestion, propose that value and its reasons once, and write\nit under `delivery` in `idd\u002Fconfig.yaml` only after approval, keeping the\nrest. Propose the `release` of `close` after the merge, ask whether it breaks\ncompatibility (a major), and cut it only with the user's OK. When they ask, use\n`work-unit-commits` to split commits, `branch-pr` to open the PR, and\n`chained-pr` when the diff exceeds the review budget. Pitch your replies at\n`team.mentorship`: `mentor` explains why, `expert` stays terse.\n\n## Memory and roadmap hand-off\n\nAfter `ospec close`, and after the delivery the user asks for, follow\n`references\u002Fmemory-loop.md`: save the `records` of `close` when memory tools\nexist, and propose the next roadmap item in any case.\n","## When to Use\n\nLoad this skill once, when the router sends a code change here (IDD is the\ndefault). It is the whole protocol — no phases, proposal or specs first.\nQuestions, explanations and read-only work need no change and do not use it.\n\nOutside a git work tree `ospec` opens no change (`not-a-git-repo`,\n`git-not-found`): set up the repository first, as `references\u002Frepository.md`\nsays. It also covers a change across several services or repositories.\n\n## The CLI decides\n\n`ospec` below means `{{ospec-cli}}`, run from the project root. Add\n`--json` to read its answer; every change command takes `--change \u003Cid>`. The\nCLI owns `idd\u002F\u003Cchange>\u002Fstate.yaml`: never write that file, and never move or\ndelete anything under `idd\u002F` by hand.\n\n- Every command that writes a change prints its next step; `ospec next\n  --change \u003Cid>` gives it in full when you resume. Its `next_step.how` tells\n  how to proceed: run the command that records the evidence or, for\n  `configure-checks`, propose a check command as `references\u002Fchecks.md` says;\n  without one no IDD change can close. Follow it.\n- An obligation is satisfied only by evidence the CLI records from a run it\n  observes (`ospec run`, `ospec check`, `ospec review`). Saying that tests\n  pass satisfies nothing.\n- The change is done when `ospec close` succeeds, and only then. Never report\n  it done, verified or archived before that.\n\n## Rules\n\n1. Read the code the request touches, then list its open facts: each behavior\n   that neither the request nor the code settles and that a careful engineer\n   would ask about. Look for defaults, invalid or edge input, error versus\n   silent handling, normalization (case, rounding, order), who may do it, and\n   whether existing behavior, output or a published API may change. Do not\n   decide them yourself.\n2. Open the change: pick a short kebab-case id from the request and record\n   the intent with `ospec record intent --change \u003Cid> --kind \u003Cbug|feature|refactor|docs> --summary \"\u003Cwhat>\" --acceptance \"\u003Cobservable result>\"`\n   plus the open facts, one `--open-fact \"\u003Cquestion>\"` each, or\n   `--no-open-facts --basis \"\u003Cwhat settles every behavior>\"` when there are\n   none. A `feature` covers the categories of the template `ospec next`\n   prints, as `references\u002Fopen-facts.md` says. Assumptions: `ospec record assumption --text`\n   (close lists them). With cross-session memory tools, recall prior\n   decisions with the `memory_query` it returns, as\n   `references\u002Fmemory-loop.md` says. Open `open-facts`: then ask all at once\n   before editing anything.\n3. If the request is materially ambiguous (no acceptance can be stated), record\n   `ospec record intent --change \u003Cid> --ambiguous --request \"\u003Crequest>\"`, ask\n   the user the question `next` returns, and resolve it with the `record\n   intent` that `next` prints, without `--ambiguous`, filling in their words\n   and the open-facts declaration of rule 2.\n4. Declare the plan before editing, on `record intent` or later with\n   `ospec signals --change \u003Cid>`: `--path \u003Cfile>` once per file you\n   expect to touch; add `--work-units \u003Cn>` when the work\n   splits into several units, `--decision \"\u003Ctext>\"` per design decision,\n   and `--operation \u003Cop>` for a destructive or irreversible operation. With\n   `docs\u002Farchitecture\u002F`, declare the ADRs and quality scenarios the change\n   bears on (`--adr \u003Cid>`, `--quality \u003Cid>`) as `references\u002Farchitecture.md`\n   says. If the plan overstated a signal, retract\n   it with `ospec record retract --change \u003Cid> --signal \u003Cid> --reason \"\u003Cwhy>\"`;\n   the CLI refuses while the diff confirms it. When its `how` offers the\n   user's answer, ask the user; never answer for them.\n5. Loop: do what the `next_step` of the last command says, repeat.\n   `resolve-gate` comes first while any gate is open: ask its\n   `pending_decision` (see Gates) and build nothing it decides until the user\n   answers. Work the obligations in the order `next` gives them:\n   - `repro-test`: write the reproduction test first and run it with\n     `ospec run --obligation repro-test --command \"\u003Ctest>\"` while it fails;\n     fix the code; run the same command again so it passes.\n   - `tdd-red-green`: the same red-then-green pair, per unit (`--unit \u003Cname>`).\n   - `contract-spec-and-test`: update the contract document and its test.\n   - `migration-compat-and-test`: run the migration test with\n     `--plan \"\u003Ccompatibility or rollback plan>\"`.\n   - `adr-impact-declaration`: `ospec record impact` for each declared ADR or\n     scenario, and their fitness checks, as `references\u002Farchitecture.md` says.\n   - `trust-review`: `ospec review start`, hand the whole returned request to\n     the read-only `review-trust` reviewer, and pass its JSON unchanged to\n     `ospec review record --result @\u003Cfile>`. Fix what it blocks on, run\n     `ospec review correct`, hand its request to the read-only\n     `review-correction` validator and pass its result to\n     `ospec review validate --result @\u003Cfile>`.\n   - `checks-pass`: `ospec check` on the current tree, or a run whose command\n     is the only check; a `tests 0` summary is not evidence. `--close`, when\n     its command has it, closes the change once ready.\n   - `living-doc`: the CLI writes the Plan and Decisions of `change.md` from\n     `--path` and `--decision`; add only what they miss. A development\n     decision stays there; it is never an ADR.\n\n   For the `suggestions`, `references` and `warnings` of `next`, read\n   `references\u002Farchitecture.md`; tell the user each warning.\n6. `ospec check` recomputes the signals from the real diff. New obligations can\n   appear: work them like the rest. One leaves `pending` only through evidence,\n   or through `ospec record withdraw --reason` when no active signal derives it.\n7. Edit text files keeping their encoding and line endings: never rewrite a\n   file through a tool or script that changes them (for example Python\n   `open()` without `encoding=` and `newline=`).\n8. When `next_step.action` is `close`, run `ospec close --change \u003Cid>`. A\n   refusal names what is still pending: settle it and close again.\n\n## Build it simply\n\nMake the smallest change that meets the acceptance. Reuse what the code\nalready has and follow its conventions. Add a layer, abstraction, dependency\nor configuration switch only for a need present in this change, and record\nthe choice with `--decision \"\u003Ctext>\"` when it is not obvious. Leave unrelated code as\nit is: report a defect outside the change instead of fixing it in passing.\n\n## Tests by risk\n\nTest what would cost something if it broke: business rules, validation,\nnon-trivial transformations and mappers, algorithms, contracts, and their\nerror, edge and regression paths. Code with nothing to get wrong (plain\naccessors, wiring without logic, declarations) needs no test of its own, and\nno coverage percentage replaces this choice.\n\n- A test must fail when the behavior is wrong: assert observable results, not\n  the calls the code makes, and check that the red run fails for the reason\n  you expect.\n- Unit tests are deterministic and offline: replace databases, external APIs\n  and infrastructure with test doubles at the boundary, and inject time and\n  randomness. An integration test, kept apart, proves the wiring.\n- A refactor writes no new test: run `ospec check` before the first edit, so\n  a recorded run shows the behavior it preserves was tested.\n\n## Gates\n\nStop for the user at the gates: end your turn with the questions, not a\nquestion tool. A `user` answer given before a turn ends is kept\n`user-unverified`.\nA skipped, empty or default answer is not the user's: end your turn with the\nquestions and wait.\n\n- `ambiguous-intent`: rule 3.\n- `open-facts`: rule 2; ask all of `pending_decision.questions` at once.\n- `irreversible-operation`: `--operation` in rule 4 or a destructive\n  statement in the diff opens it. Before any other destructive or\n  irreversible step (deleting data, rewriting published history), open it\n  yourself: `ospec record gate --change \u003Cid> --gate irreversible-operation --open --reason \"\u003Coperation>\"`.\n- `adr-amend-or-contradict`: an impact `amends`, `supersedes` or\n  `contradicts` opens it. Declare every one before asking, since the gate is\n  answered once. For a departure no declared reference covers, open it with\n  `ospec record gate --change \u003Cid> --gate adr-amend-or-contradict --open --reason \"\u003CADR>: \u003Chow the change departs from it>\"`.\n  A change that conforms opens nothing.\n- `review-limit`: `ospec check` opens it when the trust reviews run out.\n- `checks-changed`: `ospec check` opens it, running nothing, when the checks\n  or the ceremony settings (`strict_tdd`, `impact`, `contracts`) differ from\n  the change's base. The user's approval of exactly those values answers it.\n\nResolve each one with `ospec record gate --change \u003Cid> --gate \u003Cid> --resolve --answer \"\u003Ctheir words>\" --source user`,\nor `user-unverified` for a relayed answer. The CLI cannot verify who answered:\nread `references\u002Fapprovals.md` first. Only an explicit answer from the user\nresolves a gate. Outside these gates, follow the code and the evidence\nwithout asking.\n\n## Delivery\n\n`ospec close` archives the change under `idd\u002Farchive\u002F`, adds its entry to the\nchangelog `delivery.changelog` declares, and leaves the rest of the tree as it\nis. Branches, commits, pull requests and releases are the user's decision;\npropose the `team.delivery_strategy` of `next`, and the `delivery` it returns\nwhen the work starts and at close: each of its options, why it fits\n(`reasons`) and the rules to follow (`how`). With a\n`*-undeclared` suggestion, propose that value and its reasons once, and write\nit under `delivery` in `idd\u002Fconfig.yaml` only after approval, keeping the\nrest. Propose the `release` of `close` after the merge, ask whether it breaks\ncompatibility (a major), and cut it only with the user's OK. When they ask, use\n`work-unit-commits` to split commits, `branch-pr` to open the PR, and\n`chained-pr` when the diff exceeds the review budget. Pitch your replies at\n`team.mentorship`: `mentor` explains why, `expert` stays terse.\n\n## Memory and roadmap hand-off\n\nAfter `ospec close`, and after the delivery the user asks for, follow\n`references\u002Fmemory-loop.md`: save the `records` of `close` when memory tools\nexist, and propose the next roadmap item in any case.\n",{"name":5,"license":15,"metadata":24,"description":7},{"author":25,"version":26},"manuel-retamozo-garcia","1.0","main","e5fa11517ba3d433260f3077a2b8fadfad290896","skills\u002Fidd",0,"https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow","https:\u002F\u002Fospec-workflow.mretamozo.com","https:\u002F\u002Fgithub.com\u002FSnakeblack","mretamozo.com","Full Stack Developer | Lifelong learner","2026-11-15 04:24:36",{"source":8,"skillName":5,"commands":38,"prompt":48,"zipAvailable":49},[39,42,45],{"tool":40,"command":41},"npx","npx skills add Snakeblack\u002Fospec-workflow --skill idd",{"tool":43,"command":44},"bunx","bunx skills add Snakeblack\u002Fospec-workflow --skill idd",{"tool":46,"command":47},"pnpm","pnpm dlx skills add Snakeblack\u002Fospec-workflow --skill idd","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fidd\u002FSKILL.md",true,[51,57,61,65,69,73],{"path":52,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Fapprovals.md","REFERENCE","参考文档",2463,"https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fidd\u002Freferences\u002Fapprovals.md",{"path":58,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Farchitecture.md",2066,"https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fidd\u002Freferences\u002Farchitecture.md",{"path":62,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":63,"sourceHtmlUrl":64},"references\u002Fchecks.md",1581,"https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fidd\u002Freferences\u002Fchecks.md",{"path":66,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":67,"sourceHtmlUrl":68},"references\u002Fmemory-loop.md",2948,"https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fidd\u002Freferences\u002Fmemory-loop.md",{"path":70,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":71,"sourceHtmlUrl":72},"references\u002Fopen-facts.md",1937,"https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fidd\u002Freferences\u002Fopen-facts.md",{"path":74,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":75,"sourceHtmlUrl":76},"references\u002Frepository.md",2035,"https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fidd\u002Freferences\u002Frepository.md",[78,85,92,99,106,113],{"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},"01M74NGTKSC65ZZZMZH9SVP86D","work-unit-commits","Plan commits as reviewable work units. Trigger: implementation, commit splitting, chained PRs, or keeping tests and docs with code.","https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fwork-unit-commits\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":19},"01M74NGQBZ2BPFSNGREJJ2RKHG","stack-webmcp","Trigger: WebMCP, document.modelContext, web tools, AI browser agent. Web Model Context Protocol client-side tool standard.","https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fstack-webmcp\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":19},"01M74NGM8GVVSMC8T2CHGR8DFZ","stack-vite","Vite frontend build tool — configuration, plugins, asset loading","https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fstack-vite\u002FSKILL.md",[],[],{"skillNo":100,"name":101,"nameZh":6,"description":102,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":103,"categories":104,"keywords":105,"publishedAt":19},"01M74NGHF902XYK25R6AZ1XK2T","stack-starlight","Astro Starlight documentation framework — setup, sidebar navigation, frontmatter, built-in components, asides, theming, i18n, search","https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fstack-starlight\u002FSKILL.md",[],[],{"skillNo":107,"name":108,"nameZh":6,"description":109,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":110,"categories":111,"keywords":112,"publishedAt":19},"01M74NGFD2Q6GKDA7C233756X8","stack-sqlserver","Microsoft SQL Server database — indexing, RCSI, set-based operations, SARGable queries, transaction scopes","https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fstack-sqlserver\u002FSKILL.md",[],[],{"skillNo":114,"name":115,"nameZh":6,"description":116,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":117,"categories":118,"keywords":119,"publishedAt":19},"01M74NGACCR2QG9EE5HPN51QBD","stack-springboot","Spring Boot architecture patterns — REST API design, layered services, JPA, caching, Spring Security, TDD with MockMvc and Testcontainers, and pre-release verification","https:\u002F\u002Fgithub.com\u002FSnakeblack\u002Fospec-workflow\u002Fblob\u002Fmain\u002Fskills\u002Fstack-springboot\u002FSKILL.md",[],[]]