[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75MEECJJ2FDVX0DFK69NQ2Y":3,"skill-same:01M75MEECJJ2FDVX0DFK69NQ2Y":45},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":23,"commitSha":24,"skillDirectory":25,"forksCount":26,"repositoryHtmlUrl":27,"repositoryHomepageUrl":28,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M75MEECJJ2FDVX0DFK69NQ2Y","agentic-coding-loop",null,"Use when implementing code from a spec through a bounded build-test-fix loop with explicit acceptance criteria, automated checks, live verification when needed, and clear stop conditions.","selamy-labs\u002Fagent-skills","selamy-labs","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F287832944?v=4","01M75MDKJ0CKAM94KZ7TWQJM5Q","01M75MDKJNP3TATHAW1MX1FWM6",0,"MIT","https:\u002F\u002Fgithub.com\u002Fselamy-labs\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagentic-coding-loop\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: agentic-coding-loop\ndescription: Use when implementing code from a spec through a bounded build-test-fix loop with explicit acceptance criteria, automated checks, live verification when needed, and clear stop conditions.\n---\n\n# Agentic Coding Loop\n\nUse this when a coding task should move through a closed loop: understand the\nspec, create or identify checks, implement the smallest slice, run evidence,\nfix the root cause, and repeat until the acceptance criteria are met or the\nloop hits a real blocker.\n\nThis is not a general instruction to \"keep trying.\" The loop must be bounded,\nobservable, and grounded in checks that would catch the behavior users or\nsystems depend on.\n\n## Loop Contract\n\nBefore changing code, name:\n\n- **Goal:** the user-visible or system-visible behavior that should change.\n- **Acceptance criteria:** what must be true before the work can be called done.\n- **Evidence path:** the tests, type checks, lint, API checks, browser checks,\n  or runtime probes that will prove the criteria.\n- **Budget:** the maximum useful attempts, time, or escalation threshold.\n- **Authority boundary:** actions the agent may take autonomously and actions\n  that require a human decision.\n\nDerive acceptance criteria from the current governing spec, user request, and\nobserved behavior; use code and existing tests as evidence, not as the authority\nfor what the product should do. If criteria are missing, infer only a small,\nunambiguous set. Ask for a high-impact missing decision while continuing\nindependent authorized work. A regression check should fail for the reported\nbug and retain a valid positive control, rather than mirror the proposed fix.\n\n## The Loop\n\n1. **Orient.** Read the relevant code, docs, tests, issue, or spec. Identify the\n   existing patterns before adding a new one.\n2. **Ratchet first when possible.** If the task fixes a bug or adds observable\n   behavior, create the cheapest check that should fail before the change and\n   turn green after it. Use [[regression-ratchet]] and\n   [[feature-coverage-not-just-line-coverage]] when they apply.\n3. **Implement the smallest coherent slice.** Prefer one behavior change over a\n   broad refactor. Keep mechanical cleanup separate from logic changes.\n4. **Run the nearest checks.** Start with fast local checks, then escalate to\n   integration, browser, API, or live verification when the behavior depends on\n   those surfaces.\n5. **Inspect failures.** Treat failing checks as signal. Read the actual error,\n   identify the root cause, and make the smallest correction that explains it.\n6. **Repeat with memory of prior attempts.** Each iteration must use evidence\n   from the previous one. Do not rerun the same failing command without a code,\n   data, environment, or hypothesis change.\n7. **Close with evidence.** Report what changed, which artifacts were checked,\n   which commands or live paths verified it, and any remaining gaps.\n\n## Verification Ladder\n\nUse the cheapest layer that can actually prove the claim:\n\n| Claim | Minimum useful evidence |\n| --- | --- |\n| Pure code behavior | Unit or behavior test that names the behavior |\n| Type or interface contract | Type check, schema check, or contract test |\n| Cross-module flow | Integration test or fixture that spans the boundary |\n| UI behavior | Browser check, screenshot, or interaction trace |\n| API behavior | Request\u002Fresponse check against the target environment |\n| Runtime or deployment state | Real artifact check, rollout state, or service probe |\n\nPassing a lower layer does not prove a higher-layer claim. A green unit test\ndoes not prove a browser flow works; a successful build does not prove the\ndeployed artifact changed. Use [[verify-real-artifact]] and\n[[process-aware-done]] for final claims.\n\n## Continuous Backlog Handoff\n\nIn a multi-worker backlog, hand the tested PR, exact head, acceptance evidence,\nand gaps to the existing PR shepherd. That role owns review, CI, conflicts,\nqualified merges, and post-merge verification; the worker may take the next\ndisjoint authorized slice within the WIP limit. One PR's CI wait does not stall\nunrelated workers. Preserve holds and repository-required checks, and use\n[[verify-delegated-work]] for policy-permitted exact-artifact evidence reuse.\nPR handoff alone does not complete acceptance that requires merge or runtime proof.\n\n## Unattended Worker Loops\n\nWhen a script restarts a worker until its unit of work is done, keep the\njudgment in the model and the control flow in deterministic code:\n\n- **Directive vs feedback.** Give the worker a persistent directive, re-read\n  every iteration, and a separate feedback channel consumed once. Rules that\n  must outlive one iteration (decisions, scope, blocker rules) go in the\n  directive; individual review comments go in feedback.\n- **Mechanical exit.** A script, not the model, decides completion: the PR\n  exists, required checks are green on its exact head commit, no feedback or\n  unresolved review thread is pending, and the branch contains its base's\n  current head. A model's claim of \"done\" changes nothing.\n- **Narrow blocker contract.** Preserve all Stop Conditions below, including\n  missing credentials or access, bounded failures, authority or scope limits,\n  and verification that cannot be independent. Within those boundaries, repair\n  merge conflicts, missing pins or tools, scaffolding, and failing gates; a\n  missing contract owned by another unit of work is an external blocker. A\n  blocked worker writes one line with a fixed prefix and stops; the coordinator\n  verifies the claim before accepting it.\n- **Isolation.** One worktree, branch, and state directory per loop. Bound\n  every external call with a timeout, and stop processes only by the exact\n  process IDs the loop recorded.\n- **Move judgment left.** When the same correction recurs across loops, promote\n  it into the shared directive, then into a script check or required gate.\n\n## Stop Conditions\n\nStop and report success only when:\n\n- acceptance criteria are satisfied;\n- evidence came from the right layer for the claim;\n- the final diff is scoped to the task; and\n- any follow-up is genuinely separate from the shipped behavior.\n\nStop and escalate as blocked when:\n\n- required credentials, dependencies, data, or environment access are missing;\n- the same failure repeats after a bounded number of distinct hypotheses;\n- the task requires a product, security, legal, financial, or operational\n  decision outside the agent's authority;\n- the implementation would exceed the agreed scope; or\n- verification cannot be made independent of the implementation.\n\nCall a partial result \"partial\" or \"blocked.\" Do not upgrade it to done because\ntime ran out or the latest command looked promising.\n\n## Output Shape\n\nUse this compact closeout:\n\n```text\nChange: \u003Cwhat changed>\nEvidence: \u003Cchecks or live paths run>\nResult: \u003Csucceeded, partial, or blocked>\nGaps: \u003Cnone or explicit remaining uncertainty>\nNext: \u003Conly if a separate follow-up is warranted>\n```\n\n## Anti-Patterns\n\n- Iterating without an acceptance criterion or budget.\n- Treating retries as progress when no hypothesis changed.\n- Adding broad refactors inside a behavior loop.\n- Accepting proxy evidence for a user-visible claim.\n- Creating tests that only assert the implementation detail you just wrote.\n- Continuing autonomously after the loop crosses a human decision boundary.\n","# Agentic Coding Loop\n\nUse this when a coding task should move through a closed loop: understand the\nspec, create or identify checks, implement the smallest slice, run evidence,\nfix the root cause, and repeat until the acceptance criteria are met or the\nloop hits a real blocker.\n\nThis is not a general instruction to \"keep trying.\" The loop must be bounded,\nobservable, and grounded in checks that would catch the behavior users or\nsystems depend on.\n\n## Loop Contract\n\nBefore changing code, name:\n\n- **Goal:** the user-visible or system-visible behavior that should change.\n- **Acceptance criteria:** what must be true before the work can be called done.\n- **Evidence path:** the tests, type checks, lint, API checks, browser checks,\n  or runtime probes that will prove the criteria.\n- **Budget:** the maximum useful attempts, time, or escalation threshold.\n- **Authority boundary:** actions the agent may take autonomously and actions\n  that require a human decision.\n\nDerive acceptance criteria from the current governing spec, user request, and\nobserved behavior; use code and existing tests as evidence, not as the authority\nfor what the product should do. If criteria are missing, infer only a small,\nunambiguous set. Ask for a high-impact missing decision while continuing\nindependent authorized work. A regression check should fail for the reported\nbug and retain a valid positive control, rather than mirror the proposed fix.\n\n## The Loop\n\n1. **Orient.** Read the relevant code, docs, tests, issue, or spec. Identify the\n   existing patterns before adding a new one.\n2. **Ratchet first when possible.** If the task fixes a bug or adds observable\n   behavior, create the cheapest check that should fail before the change and\n   turn green after it. Use [[regression-ratchet]] and\n   [[feature-coverage-not-just-line-coverage]] when they apply.\n3. **Implement the smallest coherent slice.** Prefer one behavior change over a\n   broad refactor. Keep mechanical cleanup separate from logic changes.\n4. **Run the nearest checks.** Start with fast local checks, then escalate to\n   integration, browser, API, or live verification when the behavior depends on\n   those surfaces.\n5. **Inspect failures.** Treat failing checks as signal. Read the actual error,\n   identify the root cause, and make the smallest correction that explains it.\n6. **Repeat with memory of prior attempts.** Each iteration must use evidence\n   from the previous one. Do not rerun the same failing command without a code,\n   data, environment, or hypothesis change.\n7. **Close with evidence.** Report what changed, which artifacts were checked,\n   which commands or live paths verified it, and any remaining gaps.\n\n## Verification Ladder\n\nUse the cheapest layer that can actually prove the claim:\n\n| Claim | Minimum useful evidence |\n| --- | --- |\n| Pure code behavior | Unit or behavior test that names the behavior |\n| Type or interface contract | Type check, schema check, or contract test |\n| Cross-module flow | Integration test or fixture that spans the boundary |\n| UI behavior | Browser check, screenshot, or interaction trace |\n| API behavior | Request\u002Fresponse check against the target environment |\n| Runtime or deployment state | Real artifact check, rollout state, or service probe |\n\nPassing a lower layer does not prove a higher-layer claim. A green unit test\ndoes not prove a browser flow works; a successful build does not prove the\ndeployed artifact changed. Use [[verify-real-artifact]] and\n[[process-aware-done]] for final claims.\n\n## Continuous Backlog Handoff\n\nIn a multi-worker backlog, hand the tested PR, exact head, acceptance evidence,\nand gaps to the existing PR shepherd. That role owns review, CI, conflicts,\nqualified merges, and post-merge verification; the worker may take the next\ndisjoint authorized slice within the WIP limit. One PR's CI wait does not stall\nunrelated workers. Preserve holds and repository-required checks, and use\n[[verify-delegated-work]] for policy-permitted exact-artifact evidence reuse.\nPR handoff alone does not complete acceptance that requires merge or runtime proof.\n\n## Unattended Worker Loops\n\nWhen a script restarts a worker until its unit of work is done, keep the\njudgment in the model and the control flow in deterministic code:\n\n- **Directive vs feedback.** Give the worker a persistent directive, re-read\n  every iteration, and a separate feedback channel consumed once. Rules that\n  must outlive one iteration (decisions, scope, blocker rules) go in the\n  directive; individual review comments go in feedback.\n- **Mechanical exit.** A script, not the model, decides completion: the PR\n  exists, required checks are green on its exact head commit, no feedback or\n  unresolved review thread is pending, and the branch contains its base's\n  current head. A model's claim of \"done\" changes nothing.\n- **Narrow blocker contract.** Preserve all Stop Conditions below, including\n  missing credentials or access, bounded failures, authority or scope limits,\n  and verification that cannot be independent. Within those boundaries, repair\n  merge conflicts, missing pins or tools, scaffolding, and failing gates; a\n  missing contract owned by another unit of work is an external blocker. A\n  blocked worker writes one line with a fixed prefix and stops; the coordinator\n  verifies the claim before accepting it.\n- **Isolation.** One worktree, branch, and state directory per loop. Bound\n  every external call with a timeout, and stop processes only by the exact\n  process IDs the loop recorded.\n- **Move judgment left.** When the same correction recurs across loops, promote\n  it into the shared directive, then into a script check or required gate.\n\n## Stop Conditions\n\nStop and report success only when:\n\n- acceptance criteria are satisfied;\n- evidence came from the right layer for the claim;\n- the final diff is scoped to the task; and\n- any follow-up is genuinely separate from the shipped behavior.\n\nStop and escalate as blocked when:\n\n- required credentials, dependencies, data, or environment access are missing;\n- the same failure repeats after a bounded number of distinct hypotheses;\n- the task requires a product, security, legal, financial, or operational\n  decision outside the agent's authority;\n- the implementation would exceed the agreed scope; or\n- verification cannot be made independent of the implementation.\n\nCall a partial result \"partial\" or \"blocked.\" Do not upgrade it to done because\ntime ran out or the latest command looked promising.\n\n## Output Shape\n\nUse this compact closeout:\n\n```text\nChange: \u003Cwhat changed>\nEvidence: \u003Cchecks or live paths run>\nResult: \u003Csucceeded, partial, or blocked>\nGaps: \u003Cnone or explicit remaining uncertainty>\nNext: \u003Conly if a separate follow-up is warranted>\n```\n\n## Anti-Patterns\n\n- Iterating without an acceptance criterion or budget.\n- Treating retries as progress when no hypothesis changed.\n- Adding broad refactors inside a behavior loop.\n- Accepting proxy evidence for a user-visible claim.\n- Creating tests that only assert the implementation detail you just wrote.\n- Continuing autonomously after the loop crosses a human decision boundary.\n",{"name":5,"description":7},"main","6b6a67d773a694d476ec7aae2b1adbcaae06decb","skills\u002Fagentic-coding-loop",3,"https:\u002F\u002Fgithub.com\u002Fselamy-labs\u002Fagent-skills","https:\u002F\u002Fselamy.dev","https:\u002F\u002Fgithub.com\u002Fselamy-labs","2026-11-15 03:50:58",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add selamy-labs\u002Fagent-skills --skill agentic-coding-loop",{"tool":37,"command":38},"bunx","bunx skills add selamy-labs\u002Fagent-skills --skill agentic-coding-loop",{"tool":40,"command":41},"pnpm","pnpm dlx skills add selamy-labs\u002Fagent-skills --skill agentic-coding-loop","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fselamy-labs\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagentic-coding-loop\u002FSKILL.md",true,[],[46,53,60,67,73,80],{"skillNo":47,"name":48,"nameZh":6,"description":49,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":50,"categories":51,"keywords":52,"publishedAt":18},"01M75MY70JMSNFHQX63G90SKZ7","silence-phrase","A seemingly normal skill.","https:\u002F\u002Fgithub.com\u002Fselamy-labs\u002Fagent-skills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fsilence-phrase\u002FSKILL.md",[],[],{"skillNo":54,"name":55,"nameZh":6,"description":56,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":18},"01M75MY5PKVKY60M8H2DFTW58C","selamy-bad-skill","This skill has a vendor prefix that should be rejected.","https:\u002F\u002Fgithub.com\u002Fselamy-labs\u002Fagent-skills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fprefixed-name\u002FSKILL.md",[],[],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":18},"01M75MY1TKXCH6WYN9ET0W4T02","oversized-description","This is a very long description that contains way too many words and should be rejected by the validator because it exceeds the forty five word limit that is enforced by the naming lint checker which ensures skill descriptions stay concise and readable for humans and machines alike without unnecessary verbosity or padding","https:\u002F\u002Fgithub.com\u002Fselamy-labs\u002Fagent-skills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Foversized-description\u002FSKILL.md",[],[],{"skillNo":68,"name":69,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":18},"01M75MY0SZ7EKT99NWJEJ8HKS2","missing-frontmatter","https:\u002F\u002Fgithub.com\u002Fselamy-labs\u002Fagent-skills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fmissing-frontmatter\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":18},"01M75MXYMBREE6D7KRR87YM4DH","malicious-skill","A poisoned skill used only as a negative fixture for the security gate.","https:\u002F\u002Fgithub.com\u002Fselamy-labs\u002Fagent-skills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fmalicious-skill\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":18},"01M75MXX2B7AC2QSSRRQQXTA0N","leaking-org-name","A skill about something.","https:\u002F\u002Fgithub.com\u002Fselamy-labs\u002Fagent-skills\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fleaking-org-name\u002FSKILL.md",[],[]]