@grok/agent-sdk 0.2.29 → 0.4.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (451) hide show
  1. package/dist/channels/slack/cli-project.js +1 -1
  2. package/dist/channels/slack/slack-channel.d.ts.map +1 -1
  3. package/dist/channels/slack/slack-channel.js +27 -14
  4. package/dist/docs/404.html +2 -2
  5. package/dist/docs/assets/{app.DMrZKFjI.js → app.BSH__xV4.js} +4 -4
  6. package/dist/docs/assets/building-with-agents.md.Bo3ct-rf.js +6 -0
  7. package/dist/docs/assets/{building-with-agents.md.Bf1yH2tG.lean.js → building-with-agents.md.Bo3ct-rf.lean.js} +1 -1
  8. package/dist/docs/assets/changelog.md.b6H_ea44.js +1 -0
  9. package/dist/docs/assets/changelog.md.b6H_ea44.lean.js +1 -0
  10. package/dist/docs/assets/chunks/@localSearchIndexroot.CiRGcfZg.js +1 -0
  11. package/dist/docs/assets/chunks/{VPLocalSearchBox.DQvuw1KA.js → VPLocalSearchBox.Cadl7A0k.js} +1 -1
  12. package/dist/docs/assets/chunks/{arc.ahOG-Qza.js → arc.K7t9nNY6.js} +1 -1
  13. package/dist/docs/assets/chunks/{architectureDiagram-Q4EWVU46.BcgfYNO_.js → architectureDiagram-Q4EWVU46.SlHEknkX.js} +1 -1
  14. package/dist/docs/assets/chunks/{baseUniq.DeIYZrWh.js → baseUniq.DJLRMLkz.js} +1 -1
  15. package/dist/docs/assets/chunks/{blockDiagram-DXYQGD6D.HjJ2goMX.js → blockDiagram-DXYQGD6D.Bs6XxFpe.js} +1 -1
  16. package/dist/docs/assets/chunks/{c4Diagram-AHTNJAMY.C1QTb9GM.js → c4Diagram-AHTNJAMY.CyLg6Ovc.js} +1 -1
  17. package/dist/docs/assets/chunks/channel.CVTdqYx0.js +1 -0
  18. package/dist/docs/assets/chunks/{chunk-4BX2VUAB.UuZiDiVp.js → chunk-4BX2VUAB.DTQ4TinT.js} +1 -1
  19. package/dist/docs/assets/chunks/{chunk-4TB4RGXK.5A3YmkhI.js → chunk-4TB4RGXK.Es84IgcH.js} +1 -1
  20. package/dist/docs/assets/chunks/{chunk-55IACEB6.B4uf1yYV.js → chunk-55IACEB6.Clm3cGPQ.js} +1 -1
  21. package/dist/docs/assets/chunks/{chunk-EDXVE4YY.CoXBce8t.js → chunk-EDXVE4YY.DE4CRkS5.js} +1 -1
  22. package/dist/docs/assets/chunks/{chunk-FMBD7UC4.DnC5NO4e.js → chunk-FMBD7UC4.Bf8HCvqp.js} +1 -1
  23. package/dist/docs/assets/chunks/{chunk-OYMX7WX6.CzfHCeIi.js → chunk-OYMX7WX6.eub-7nSx.js} +1 -1
  24. package/dist/docs/assets/chunks/{chunk-QZHKN3VN.C6tcwkVd.js → chunk-QZHKN3VN.WyP9VAS_.js} +1 -1
  25. package/dist/docs/assets/chunks/{chunk-YZCP3GAM._oboyZ1A.js → chunk-YZCP3GAM.LO9ohhDE.js} +1 -1
  26. package/dist/docs/assets/chunks/classDiagram-6PBFFD2Q.CsJJyx6-.js +1 -0
  27. package/dist/docs/assets/chunks/classDiagram-v2-HSJHXN6E.CsJJyx6-.js +1 -0
  28. package/dist/docs/assets/chunks/clone.Dk8KvtOB.js +1 -0
  29. package/dist/docs/assets/chunks/{cose-bilkent-S5V4N54A.BfPVKeId.js → cose-bilkent-S5V4N54A.D5za6aH0.js} +1 -1
  30. package/dist/docs/assets/chunks/{dagre-KV5264BT.CLMLc0lD.js → dagre-KV5264BT.Dq1daUru.js} +1 -1
  31. package/dist/docs/assets/chunks/{diagram-5BDNPKRD.C7OF7UfH.js → diagram-5BDNPKRD.j1YL5BKS.js} +1 -1
  32. package/dist/docs/assets/chunks/{diagram-G4DWMVQ6.Dpao0Bl1.js → diagram-G4DWMVQ6.Bi1aLdhC.js} +1 -1
  33. package/dist/docs/assets/chunks/{diagram-MMDJMWI5.DLnM9eLp.js → diagram-MMDJMWI5.BOzMEkAr.js} +1 -1
  34. package/dist/docs/assets/chunks/{diagram-TYMM5635.BcpVgA_M.js → diagram-TYMM5635.Dx2FVhCr.js} +1 -1
  35. package/dist/docs/assets/chunks/{erDiagram-SMLLAGMA.CsL5JuCV.js → erDiagram-SMLLAGMA.DI4TNkye.js} +1 -1
  36. package/dist/docs/assets/chunks/{flowDiagram-DWJPFMVM.MFihq9AF.js → flowDiagram-DWJPFMVM.C58Z9uPK.js} +1 -1
  37. package/dist/docs/assets/chunks/{ganttDiagram-T4ZO3ILL.CJeLodyG.js → ganttDiagram-T4ZO3ILL.DXSpW809.js} +1 -1
  38. package/dist/docs/assets/chunks/{gitGraphDiagram-UUTBAWPF.1HhlE45Z.js → gitGraphDiagram-UUTBAWPF.B955oaFl.js} +1 -1
  39. package/dist/docs/assets/chunks/{graph.BS02KkWA.js → graph.BQXhKtxS.js} +1 -1
  40. package/dist/docs/assets/chunks/{infoDiagram-42DDH7IO.fp5gvytQ.js → infoDiagram-42DDH7IO.Dozj_-Gm.js} +1 -1
  41. package/dist/docs/assets/chunks/{ishikawaDiagram-UXIWVN3A.C6Wk1V4I.js → ishikawaDiagram-UXIWVN3A.D2140OzH.js} +1 -1
  42. package/dist/docs/assets/chunks/{journeyDiagram-VCZTEJTY.qJt4uJ_H.js → journeyDiagram-VCZTEJTY.DrqrMhCk.js} +1 -1
  43. package/dist/docs/assets/chunks/{kanban-definition-6JOO6SKY.z2tG9kly.js → kanban-definition-6JOO6SKY.Dv-LTrlw.js} +1 -1
  44. package/dist/docs/assets/chunks/{layout.BPCkdejr.js → layout.CGtHp49X.js} +1 -1
  45. package/dist/docs/assets/chunks/{linear.CL3TC3u7.js → linear.CHZLYB52.js} +1 -1
  46. package/dist/docs/assets/chunks/{min.SPrSp704.js → min.D4Vt3eIy.js} +1 -1
  47. package/dist/docs/assets/chunks/{mindmap-definition-QFDTVHPH.DgrQRFZW.js → mindmap-definition-QFDTVHPH.BBfpYTea.js} +1 -1
  48. package/dist/docs/assets/chunks/{pieDiagram-DEJITSTG.Ba7yxiW2.js → pieDiagram-DEJITSTG.BVYvElmY.js} +1 -1
  49. package/dist/docs/assets/chunks/{quadrantDiagram-34T5L4WZ.Bvskad4p.js → quadrantDiagram-34T5L4WZ.C6Gc2mdY.js} +1 -1
  50. package/dist/docs/assets/chunks/{requirementDiagram-MS252O5E.Ce70ugV9.js → requirementDiagram-MS252O5E.DWa8b4cn.js} +1 -1
  51. package/dist/docs/assets/chunks/{sankeyDiagram-XADWPNL6.BJ5OEULO.js → sankeyDiagram-XADWPNL6.Bi-g4ajv.js} +1 -1
  52. package/dist/docs/assets/chunks/{sequenceDiagram-FGHM5R23.C8a1hDs6.js → sequenceDiagram-FGHM5R23.DlU3vVN9.js} +1 -1
  53. package/dist/docs/assets/chunks/{stateDiagram-FHFEXIEX.D2D8A5CK.js → stateDiagram-FHFEXIEX.CeqJRqnU.js} +1 -1
  54. package/dist/docs/assets/chunks/stateDiagram-v2-QKLJ7IA2.DE0BnUZh.js +1 -0
  55. package/dist/docs/assets/chunks/{theme.DRciHmS1.js → theme.DGq7D4tt.js} +2 -2
  56. package/dist/docs/assets/chunks/{timeline-definition-GMOUNBTQ.C4ePrY8-.js → timeline-definition-GMOUNBTQ.Cqf2kjq9.js} +1 -1
  57. package/dist/docs/assets/chunks/{vennDiagram-DHZGUBPP.Dzso6dEA.js → vennDiagram-DHZGUBPP.BeOpE_lM.js} +1 -1
  58. package/dist/docs/assets/chunks/{wardley-RL74JXVD.djgXR0ar.js → wardley-RL74JXVD.FMQXA-6q.js} +1 -1
  59. package/dist/docs/assets/chunks/{wardleyDiagram-NUSXRM2D.Bf2ICxjh.js → wardleyDiagram-NUSXRM2D.BXS65dAC.js} +1 -1
  60. package/dist/docs/assets/chunks/{xychartDiagram-5P7HB3ND.DEH4BcVJ.js → xychartDiagram-5P7HB3ND.haEiSxqc.js} +1 -1
  61. package/dist/docs/assets/deployment.md.DJtuDVLh.js +31 -0
  62. package/dist/docs/assets/{deployment.md.KW0Ni_rY.lean.js → deployment.md.DJtuDVLh.lean.js} +1 -1
  63. package/dist/docs/assets/evals.md.Dne6VRmZ.js +49 -0
  64. package/dist/docs/assets/evals.md.Dne6VRmZ.lean.js +1 -0
  65. package/dist/docs/assets/{guides_agent-to-agent.md.D1D5f6Jc.js → guides_agent-to-agent.md.DcB3nhOz.js} +8 -8
  66. package/dist/docs/assets/guides_agent-to-agent.md.DcB3nhOz.lean.js +1 -0
  67. package/dist/docs/assets/{guides_bitbucket.md.DkjkBRU8.js → guides_bitbucket.md.BiA6ZMFW.js} +27 -27
  68. package/dist/docs/assets/{guides_bitbucket.md.DkjkBRU8.lean.js → guides_bitbucket.md.BiA6ZMFW.lean.js} +1 -1
  69. package/dist/docs/assets/guides_cloud-agents.md.CiXAW6on.js +9 -0
  70. package/dist/docs/assets/{guides_cloud-agents.md.Ceyjv0ul.lean.js → guides_cloud-agents.md.CiXAW6on.lean.js} +1 -1
  71. package/dist/docs/assets/guides_convert-automation.md.BvTxsJ8J.js +26 -0
  72. package/dist/docs/assets/{guides_convert-automation.md.BjraCvok.lean.js → guides_convert-automation.md.BvTxsJ8J.lean.js} +1 -1
  73. package/dist/docs/assets/{guides_github.md.CgN5sNbW.js → guides_github.md.CcuZ7keS.js} +11 -19
  74. package/dist/docs/assets/{guides_github.md.CgN5sNbW.lean.js → guides_github.md.CcuZ7keS.lean.js} +1 -1
  75. package/dist/docs/assets/{guides_gitlab.md.DCE0p7_w.js → guides_gitlab.md.DcKusOcU.js} +63 -73
  76. package/dist/docs/assets/{guides_gitlab.md.DCE0p7_w.lean.js → guides_gitlab.md.DcKusOcU.lean.js} +1 -1
  77. package/dist/docs/assets/guides_improve.md.o3WOGR7n.js +16 -0
  78. package/dist/docs/assets/{guides_improve.md.BFYtrtLE.lean.js → guides_improve.md.o3WOGR7n.lean.js} +1 -1
  79. package/dist/docs/assets/{guides_jev.md.CEqLgjj9.js → guides_jev.md.DGCzkMUN.js} +20 -60
  80. package/dist/docs/assets/{guides_jev.md.CEqLgjj9.lean.js → guides_jev.md.DGCzkMUN.lean.js} +1 -1
  81. package/dist/docs/assets/{guides_mcp-oauth.md.DrGY92u4.js → guides_mcp-oauth.md.GVPMKcYl.js} +3 -5
  82. package/dist/docs/assets/{guides_mcp-oauth.md.DrGY92u4.lean.js → guides_mcp-oauth.md.GVPMKcYl.lean.js} +1 -1
  83. package/dist/docs/assets/guides_migrate.md.B8CZxnnX.js +3 -0
  84. package/dist/docs/assets/{guides_migrate.md.RS9Eu-Xh.lean.js → guides_migrate.md.B8CZxnnX.lean.js} +1 -1
  85. package/dist/docs/assets/guides_opentelemetry.md.BRSktBHt.js +26 -0
  86. package/dist/docs/assets/guides_opentelemetry.md.BRSktBHt.lean.js +1 -0
  87. package/dist/docs/assets/guides_pagerduty.md.pMXUkZZc.js +61 -0
  88. package/dist/docs/assets/guides_pagerduty.md.pMXUkZZc.lean.js +1 -0
  89. package/dist/docs/assets/guides_slack.md.D7Rwyva4.js +82 -0
  90. package/dist/docs/assets/{guides_slack.md.BwXHURvY.lean.js → guides_slack.md.D7Rwyva4.lean.js} +1 -1
  91. package/dist/docs/assets/{guides_webhooks.md.BJk63ZTs.js → guides_webhooks.md.B5DXt2v1.js} +7 -7
  92. package/dist/docs/assets/{guides_webhooks.md.BJk63ZTs.lean.js → guides_webhooks.md.B5DXt2v1.lean.js} +1 -1
  93. package/dist/docs/assets/hillclimbing.md.M51bF-oT.js +5 -0
  94. package/dist/docs/assets/hillclimbing.md.M51bF-oT.lean.js +1 -0
  95. package/dist/docs/assets/index.md.DnJPj_B6.js +14 -0
  96. package/dist/docs/assets/index.md.DnJPj_B6.lean.js +1 -0
  97. package/dist/docs/assets/quickstart.md.r-oE8nkX.js +113 -0
  98. package/dist/docs/assets/{quickstart.md.CljDaJBi.lean.js → quickstart.md.r-oE8nkX.lean.js} +1 -1
  99. package/dist/docs/assets/reference_access.md.C5lblfUC.js +1 -0
  100. package/dist/docs/assets/reference_access.md.C5lblfUC.lean.js +1 -0
  101. package/dist/docs/assets/reference_agent-config.md.BSNp2uJ7.js +27 -0
  102. package/dist/docs/assets/{reference_agent-config.md.Ck29qYNj.lean.js → reference_agent-config.md.BSNp2uJ7.lean.js} +1 -1
  103. package/dist/docs/assets/reference_artifacts.md.BJpAI7h9.js +17 -0
  104. package/dist/docs/assets/{reference_artifacts.md.C-hKR2hp.lean.js → reference_artifacts.md.BJpAI7h9.lean.js} +1 -1
  105. package/dist/docs/assets/reference_channels.md.CPWOpdsC.js +21 -0
  106. package/dist/docs/assets/{reference_channels.md.DsJfGSBJ.lean.js → reference_channels.md.CPWOpdsC.lean.js} +1 -1
  107. package/dist/docs/assets/reference_cli.md.DHm6DmYk.js +83 -0
  108. package/dist/docs/assets/reference_cli.md.DHm6DmYk.lean.js +1 -0
  109. package/dist/docs/assets/reference_connections.md.DoY4vrDH.js +54 -0
  110. package/dist/docs/assets/{reference_connections.md.BoAotr7_.lean.js → reference_connections.md.DoY4vrDH.lean.js} +1 -1
  111. package/dist/docs/assets/reference_evals.md.DgfUtfIc.js +69 -0
  112. package/dist/docs/assets/{reference_evals.md.2BXeMiO2.lean.js → reference_evals.md.DgfUtfIc.lean.js} +1 -1
  113. package/dist/docs/assets/{reference_extensions.md.DI5X5x8p.js → reference_extensions.md.DHbW6eOf.js} +6 -8
  114. package/dist/docs/assets/{reference_extensions.md.DI5X5x8p.lean.js → reference_extensions.md.DHbW6eOf.lean.js} +1 -1
  115. package/dist/docs/assets/reference_hooks.md.vbIM394-.js +39 -0
  116. package/dist/docs/assets/reference_hooks.md.vbIM394-.lean.js +1 -0
  117. package/dist/docs/assets/reference_http-api.md.C7Cb0r2J.js +7 -0
  118. package/dist/docs/assets/{reference_http-api.md._RWELr3d.lean.js → reference_http-api.md.C7Cb0r2J.lean.js} +1 -1
  119. package/dist/docs/assets/reference_instructions.md.CV6Qq0UX.js +14 -0
  120. package/dist/docs/assets/reference_instructions.md.CV6Qq0UX.lean.js +1 -0
  121. package/dist/docs/assets/reference_limits.md.CbdtFr4T.js +8 -0
  122. package/dist/docs/assets/reference_limits.md.CbdtFr4T.lean.js +1 -0
  123. package/dist/docs/assets/reference_memory.md.CdoJISM-.js +3 -0
  124. package/dist/docs/assets/{reference_memory.md.Dd7cyV9V.lean.js → reference_memory.md.CdoJISM-.lean.js} +1 -1
  125. package/dist/docs/assets/reference_playground.md.DIUsRAcy.js +1 -0
  126. package/dist/docs/assets/{reference_playground.md.CQiJk9CM.lean.js → reference_playground.md.DIUsRAcy.lean.js} +1 -1
  127. package/dist/docs/assets/reference_project-layout.md.9DiKoE16.js +26 -0
  128. package/dist/docs/assets/reference_project-layout.md.9DiKoE16.lean.js +1 -0
  129. package/dist/docs/assets/reference_schedules.md.Dd8_d0WG.js +45 -0
  130. package/dist/docs/assets/{reference_schedules.md.RC47aFm4.lean.js → reference_schedules.md.Dd8_d0WG.lean.js} +1 -1
  131. package/dist/docs/assets/reference_sessions.md.NXVQt9mP.js +22 -0
  132. package/dist/docs/assets/{reference_sessions.md.C7KFhWXV.lean.js → reference_sessions.md.NXVQt9mP.lean.js} +1 -1
  133. package/dist/docs/assets/{reference_skills.md.Dej2zQxi.js → reference_skills.md.DzCPNPkG.js} +4 -4
  134. package/dist/docs/assets/{reference_skills.md.Dej2zQxi.lean.js → reference_skills.md.DzCPNPkG.lean.js} +1 -1
  135. package/dist/docs/assets/reference_subagents.md.Dvpq2sUu.js +10 -0
  136. package/dist/docs/assets/{reference_subagents.md.CAhvW4yS.lean.js → reference_subagents.md.Dvpq2sUu.lean.js} +1 -1
  137. package/dist/docs/assets/reference_tools.md.B2OaR1Nb.js +62 -0
  138. package/dist/docs/assets/{reference_tools.md.DQr7Fnhu.lean.js → reference_tools.md.B2OaR1Nb.lean.js} +1 -1
  139. package/dist/docs/assets/templates_agentic-owners.md.B6ZmKAzT.js +13 -0
  140. package/dist/docs/assets/{templates_agentic-owners.md.Dh2nDSqg.lean.js → templates_agentic-owners.md.B6ZmKAzT.lean.js} +1 -1
  141. package/dist/docs/assets/templates_reviews.md.dr9Xv_KM.js +10 -0
  142. package/dist/docs/assets/templates_reviews.md.dr9Xv_KM.lean.js +1 -0
  143. package/dist/docs/assets/troubleshooting.md.DfKKvW1R.js +3 -0
  144. package/dist/docs/assets/{troubleshooting.md.CGOdNVAL.lean.js → troubleshooting.md.DfKKvW1R.lean.js} +1 -1
  145. package/dist/docs/building-with-agents.html +41 -44
  146. package/dist/docs/building-with-agents.md +11 -104
  147. package/dist/docs/changelog.html +37 -37
  148. package/dist/docs/changelog.md +14 -145
  149. package/dist/docs/deployment.html +54 -54
  150. package/dist/docs/deployment.md +97 -270
  151. package/dist/docs/evals.html +52 -71
  152. package/dist/docs/evals.md +33 -129
  153. package/dist/docs/guides/agent-to-agent.html +43 -43
  154. package/dist/docs/guides/agent-to-agent.md +29 -64
  155. package/dist/docs/guides/bitbucket.html +62 -62
  156. package/dist/docs/guides/bitbucket.md +73 -134
  157. package/dist/docs/guides/cloud-agents.html +37 -37
  158. package/dist/docs/guides/cloud-agents.md +23 -53
  159. package/dist/docs/guides/convert-automation.html +39 -56
  160. package/dist/docs/guides/convert-automation.md +31 -115
  161. package/dist/docs/guides/github.html +46 -54
  162. package/dist/docs/guides/github.md +49 -131
  163. package/dist/docs/guides/gitlab.html +98 -108
  164. package/dist/docs/guides/gitlab.md +104 -184
  165. package/dist/docs/guides/improve.html +38 -43
  166. package/dist/docs/guides/improve.md +24 -90
  167. package/dist/docs/guides/jev.html +55 -95
  168. package/dist/docs/guides/jev.md +53 -160
  169. package/dist/docs/guides/mcp-oauth.html +38 -40
  170. package/dist/docs/guides/mcp-oauth.md +33 -67
  171. package/dist/docs/guides/migrate.html +38 -43
  172. package/dist/docs/guides/migrate.md +56 -130
  173. package/dist/docs/guides/opentelemetry.html +41 -43
  174. package/dist/docs/guides/opentelemetry.md +32 -27
  175. package/dist/docs/guides/pagerduty.html +50 -49
  176. package/dist/docs/guides/pagerduty.md +65 -199
  177. package/dist/docs/guides/slack.html +51 -57
  178. package/dist/docs/guides/slack.md +76 -204
  179. package/dist/docs/guides/webhooks.html +42 -42
  180. package/dist/docs/guides/webhooks.md +64 -146
  181. package/dist/docs/hashmap.json +1 -1
  182. package/dist/docs/hillclimbing.html +40 -42
  183. package/dist/docs/hillclimbing.md +40 -102
  184. package/dist/docs/index.html +49 -61
  185. package/dist/docs/index.md +42 -82
  186. package/dist/docs/llms-full.txt +2272 -5711
  187. package/dist/docs/llms.txt +9 -14
  188. package/dist/docs/quickstart.html +95 -152
  189. package/dist/docs/quickstart.md +91 -180
  190. package/dist/docs/reference/access.html +36 -36
  191. package/dist/docs/reference/access.md +7 -12
  192. package/dist/docs/reference/agent-config.html +45 -58
  193. package/dist/docs/reference/agent-config.md +92 -176
  194. package/dist/docs/reference/artifacts.html +38 -39
  195. package/dist/docs/reference/artifacts.md +38 -62
  196. package/dist/docs/reference/channels.html +38 -39
  197. package/dist/docs/reference/channels.md +102 -151
  198. package/dist/docs/reference/cli.html +68 -84
  199. package/dist/docs/reference/cli.md +232 -697
  200. package/dist/docs/reference/connections.html +45 -57
  201. package/dist/docs/reference/connections.md +83 -164
  202. package/dist/docs/reference/evals.html +47 -54
  203. package/dist/docs/reference/evals.md +75 -141
  204. package/dist/docs/reference/extensions.html +41 -43
  205. package/dist/docs/reference/extensions.md +37 -70
  206. package/dist/docs/reference/hooks.html +40 -52
  207. package/dist/docs/reference/hooks.md +62 -133
  208. package/dist/docs/reference/http-api.html +38 -42
  209. package/dist/docs/reference/http-api.md +66 -118
  210. package/dist/docs/reference/instructions.html +39 -39
  211. package/dist/docs/reference/instructions.md +19 -29
  212. package/dist/docs/reference/limits.html +38 -38
  213. package/dist/docs/reference/limits.md +9 -21
  214. package/dist/docs/reference/memory.html +37 -38
  215. package/dist/docs/reference/memory.md +38 -75
  216. package/dist/docs/reference/playground.html +36 -36
  217. package/dist/docs/reference/playground.md +14 -22
  218. package/dist/docs/reference/project-layout.html +49 -48
  219. package/dist/docs/reference/project-layout.md +38 -84
  220. package/dist/docs/reference/schedules.html +44 -46
  221. package/dist/docs/reference/schedules.md +59 -90
  222. package/dist/docs/reference/sessions.html +37 -37
  223. package/dist/docs/reference/sessions.md +66 -113
  224. package/dist/docs/reference/skills.html +39 -39
  225. package/dist/docs/reference/skills.md +11 -31
  226. package/dist/docs/reference/subagents.html +37 -37
  227. package/dist/docs/reference/subagents.md +17 -31
  228. package/dist/docs/reference/tools.html +47 -64
  229. package/dist/docs/reference/tools.md +48 -130
  230. package/dist/docs/templates/agentic-owners.html +38 -38
  231. package/dist/docs/templates/agentic-owners.md +50 -74
  232. package/dist/docs/templates/reviews.html +69 -0
  233. package/dist/docs/templates/reviews.md +71 -0
  234. package/dist/docs/troubleshooting.html +38 -36
  235. package/dist/docs/troubleshooting.md +48 -101
  236. package/dist/filesystem.d.ts +7 -320
  237. package/dist/filesystem.d.ts.map +1 -1
  238. package/dist/filesystem.js +10 -691
  239. package/dist/index.d.ts +1 -1
  240. package/dist/index.d.ts.map +1 -1
  241. package/dist/index.js +1 -1
  242. package/dist/internal/authored-alias-hooks.d.ts +6 -0
  243. package/dist/internal/authored-alias-hooks.d.ts.map +1 -1
  244. package/dist/internal/authored-alias-hooks.js +12 -0
  245. package/dist/internal/authored-loaders.js +10 -1
  246. package/dist/internal/cli/commands/scaffold.js +1 -1
  247. package/dist/internal/cli-deploy.js +1 -1
  248. package/dist/internal/cli-docs.js +1 -1
  249. package/dist/internal/cli-images.js +1 -1
  250. package/dist/internal/cli-skills.js +1 -1
  251. package/dist/internal/continuation-session-ids.d.ts +15 -0
  252. package/dist/internal/continuation-session-ids.d.ts.map +1 -0
  253. package/dist/internal/continuation-session-ids.js +43 -0
  254. package/dist/internal/cursor/credentials.js +1 -1
  255. package/dist/internal/deploy-manifest.d.ts +1 -0
  256. package/dist/internal/deploy-manifest.d.ts.map +1 -1
  257. package/dist/internal/deploy-manifest.js +0 -1
  258. package/dist/internal/discovery/agent-config.d.ts +2 -6
  259. package/dist/internal/discovery/agent-config.d.ts.map +1 -1
  260. package/dist/internal/discovery/agent-config.js +8 -25
  261. package/dist/internal/discovery/agent.d.ts.map +1 -1
  262. package/dist/internal/discovery/agent.js +2 -3
  263. package/dist/internal/discovery/project.d.ts.map +1 -1
  264. package/dist/internal/discovery/project.js +15 -0
  265. package/dist/internal/filesystem/walk.d.ts +1 -5
  266. package/dist/internal/filesystem/walk.d.ts.map +1 -1
  267. package/dist/internal/filesystem/walk.js +1 -9
  268. package/dist/internal/hosted-catch-protocol.d.ts +3 -3
  269. package/dist/internal/hosted-catch-protocol.js +3 -3
  270. package/dist/internal/hosted-fetch.d.ts +7 -0
  271. package/dist/internal/hosted-fetch.d.ts.map +1 -0
  272. package/dist/internal/hosted-fetch.js +26 -0
  273. package/dist/internal/init-project.d.ts.map +1 -1
  274. package/dist/internal/init-project.js +7 -5
  275. package/dist/internal/install-cursor-skills.d.ts +5 -6
  276. package/dist/internal/install-cursor-skills.d.ts.map +1 -1
  277. package/dist/internal/install-cursor-skills.js +19 -4
  278. package/dist/internal/mcp-endpoint.js +4 -4
  279. package/dist/internal/model-config.d.ts.map +1 -1
  280. package/dist/internal/model-config.js +11 -4
  281. package/dist/internal/platform-timers.d.ts +3 -7
  282. package/dist/internal/platform-timers.d.ts.map +1 -1
  283. package/dist/internal/platform-timers.js +10 -51
  284. package/dist/internal/reminder-runner.js +1 -1
  285. package/dist/internal/server.d.ts.map +1 -1
  286. package/dist/internal/server.js +6 -1
  287. package/dist/internal/session-engine.d.ts +3 -0
  288. package/dist/internal/session-engine.d.ts.map +1 -1
  289. package/dist/internal/session-engine.js +13 -4
  290. package/dist/internal/shallow-clone.js +1 -1
  291. package/dist/playground/assets/index-Bno25Cg0.css +1 -0
  292. package/dist/playground/assets/index-SGKsaznh.js +62 -0
  293. package/dist/playground/index.html +2 -2
  294. package/dist/types.d.ts +6 -17
  295. package/dist/types.d.ts.map +1 -1
  296. package/dist/types.js +2 -2
  297. package/docs/README.md +42 -82
  298. package/docs/building-with-agents.md +11 -104
  299. package/docs/changelog.md +15 -146
  300. package/docs/deployment.md +97 -270
  301. package/docs/evals.md +34 -130
  302. package/docs/guides/agent-to-agent.md +29 -64
  303. package/docs/guides/bitbucket.md +73 -134
  304. package/docs/guides/cloud-agents.md +23 -53
  305. package/docs/guides/convert-automation.md +31 -115
  306. package/docs/guides/github.md +49 -131
  307. package/docs/guides/gitlab.md +104 -184
  308. package/docs/guides/improve.md +24 -90
  309. package/docs/guides/jev.md +53 -160
  310. package/docs/guides/mcp-oauth.md +33 -67
  311. package/docs/guides/migrate.md +56 -130
  312. package/docs/guides/opentelemetry.md +37 -27
  313. package/docs/guides/pagerduty.md +68 -200
  314. package/docs/guides/slack.md +76 -204
  315. package/docs/guides/webhooks.md +64 -146
  316. package/docs/hillclimbing.md +41 -103
  317. package/docs/quickstart.md +91 -180
  318. package/docs/reference/access.md +7 -12
  319. package/docs/reference/agent-config.md +92 -176
  320. package/docs/reference/artifacts.md +38 -62
  321. package/docs/reference/channels.md +102 -151
  322. package/docs/reference/cli.md +232 -697
  323. package/docs/reference/connections.md +83 -164
  324. package/docs/reference/evals.md +75 -141
  325. package/docs/reference/extensions.md +37 -70
  326. package/docs/reference/hooks.md +61 -132
  327. package/docs/reference/http-api.md +66 -118
  328. package/docs/reference/instructions.md +20 -30
  329. package/docs/reference/limits.md +10 -22
  330. package/docs/reference/memory.md +38 -75
  331. package/docs/reference/playground.md +14 -22
  332. package/docs/reference/project-layout.md +38 -84
  333. package/docs/reference/schedules.md +59 -90
  334. package/docs/reference/sessions.md +66 -113
  335. package/docs/reference/skills.md +12 -32
  336. package/docs/reference/subagents.md +17 -31
  337. package/docs/reference/tools.md +48 -130
  338. package/docs/templates/agentic-owners.md +50 -74
  339. package/docs/templates/reviews.md +76 -0
  340. package/docs/troubleshooting.md +48 -101
  341. package/package.json +4 -4
  342. package/skills/cli/SKILL.md +1 -1
  343. package/skills/create-agent/SKILL.md +61 -7
  344. package/skills/debug/SKILL.md +1 -1
  345. package/skills/deploy/SKILL.md +3 -3
  346. package/skills/evals/SKILL.md +1 -1
  347. package/skills/framework-map/SKILL.md +2 -2
  348. package/skills/github/SKILL.md +1 -1
  349. package/skills/hillclimb/SKILL.md +1 -1
  350. package/skills/mcp-auth/SKILL.md +1 -1
  351. package/skills/migrate-v2/SKILL.md +2 -3
  352. package/skills/otel/SKILL.md +1 -1
  353. package/skills/setup-slack/SKILL.md +2 -2
  354. package/dist/docs/assets/building-with-agents.md.Bf1yH2tG.js +0 -9
  355. package/dist/docs/assets/changelog.md.DrLFQUlu.js +0 -1
  356. package/dist/docs/assets/changelog.md.DrLFQUlu.lean.js +0 -1
  357. package/dist/docs/assets/chunks/@localSearchIndexroot.Dpd_wxVU.js +0 -1
  358. package/dist/docs/assets/chunks/channel._fZX_hvU.js +0 -1
  359. package/dist/docs/assets/chunks/classDiagram-6PBFFD2Q.eClS1Q0E.js +0 -1
  360. package/dist/docs/assets/chunks/classDiagram-v2-HSJHXN6E.eClS1Q0E.js +0 -1
  361. package/dist/docs/assets/chunks/clone.W1Phdw9P.js +0 -1
  362. package/dist/docs/assets/chunks/stateDiagram-v2-QKLJ7IA2.77yf4kRr.js +0 -1
  363. package/dist/docs/assets/deployment.md.KW0Ni_rY.js +0 -31
  364. package/dist/docs/assets/evals.md.DiljU__Z.js +0 -68
  365. package/dist/docs/assets/evals.md.DiljU__Z.lean.js +0 -1
  366. package/dist/docs/assets/guides_agent-to-agent.md.D1D5f6Jc.lean.js +0 -1
  367. package/dist/docs/assets/guides_cloud-agents.md.Ceyjv0ul.js +0 -9
  368. package/dist/docs/assets/guides_convert-automation.md.BjraCvok.js +0 -43
  369. package/dist/docs/assets/guides_hooks.md.BjeEiM7s.js +0 -54
  370. package/dist/docs/assets/guides_hooks.md.BjeEiM7s.lean.js +0 -1
  371. package/dist/docs/assets/guides_improve.md.BFYtrtLE.js +0 -21
  372. package/dist/docs/assets/guides_migrate.md.RS9Eu-Xh.js +0 -8
  373. package/dist/docs/assets/guides_opentelemetry.md.CoZ1kNyd.js +0 -28
  374. package/dist/docs/assets/guides_opentelemetry.md.CoZ1kNyd.lean.js +0 -1
  375. package/dist/docs/assets/guides_pagerduty.md.mnKXe9U0.js +0 -60
  376. package/dist/docs/assets/guides_pagerduty.md.mnKXe9U0.lean.js +0 -1
  377. package/dist/docs/assets/guides_slack.md.BwXHURvY.js +0 -88
  378. package/dist/docs/assets/hillclimbing.md.By61f3uq.js +0 -7
  379. package/dist/docs/assets/hillclimbing.md.By61f3uq.lean.js +0 -1
  380. package/dist/docs/assets/index.md.CAruw_rp.js +0 -26
  381. package/dist/docs/assets/index.md.CAruw_rp.lean.js +0 -1
  382. package/dist/docs/assets/quickstart.md.CljDaJBi.js +0 -170
  383. package/dist/docs/assets/reference_access.md.CEtibb4i.js +0 -1
  384. package/dist/docs/assets/reference_access.md.CEtibb4i.lean.js +0 -1
  385. package/dist/docs/assets/reference_agent-config.md.Ck29qYNj.js +0 -40
  386. package/dist/docs/assets/reference_artifacts.md.C-hKR2hp.js +0 -18
  387. package/dist/docs/assets/reference_channels.md.DsJfGSBJ.js +0 -22
  388. package/dist/docs/assets/reference_cli.md.BkdLRbXg.js +0 -99
  389. package/dist/docs/assets/reference_cli.md.BkdLRbXg.lean.js +0 -1
  390. package/dist/docs/assets/reference_connections.md.BoAotr7_.js +0 -66
  391. package/dist/docs/assets/reference_evals.md.2BXeMiO2.js +0 -76
  392. package/dist/docs/assets/reference_hooks.md.B-bwJuiG.js +0 -51
  393. package/dist/docs/assets/reference_hooks.md.B-bwJuiG.lean.js +0 -1
  394. package/dist/docs/assets/reference_http-api.md._RWELr3d.js +0 -11
  395. package/dist/docs/assets/reference_instructions.md.yL2hXLnl.js +0 -14
  396. package/dist/docs/assets/reference_instructions.md.yL2hXLnl.lean.js +0 -1
  397. package/dist/docs/assets/reference_limits.md.BCCnvt3o.js +0 -8
  398. package/dist/docs/assets/reference_limits.md.BCCnvt3o.lean.js +0 -1
  399. package/dist/docs/assets/reference_memory.md.Dd7cyV9V.js +0 -4
  400. package/dist/docs/assets/reference_playground.md.CQiJk9CM.js +0 -1
  401. package/dist/docs/assets/reference_project-layout.md.BamcPDUn.js +0 -25
  402. package/dist/docs/assets/reference_project-layout.md.BamcPDUn.lean.js +0 -1
  403. package/dist/docs/assets/reference_prompt.md.BMjm1Jp3.js +0 -9
  404. package/dist/docs/assets/reference_prompt.md.BMjm1Jp3.lean.js +0 -1
  405. package/dist/docs/assets/reference_schedules.md.RC47aFm4.js +0 -47
  406. package/dist/docs/assets/reference_sessions.md.C7KFhWXV.js +0 -22
  407. package/dist/docs/assets/reference_subagents.md.CAhvW4yS.js +0 -10
  408. package/dist/docs/assets/reference_tools.md.DQr7Fnhu.js +0 -79
  409. package/dist/docs/assets/templates_agentic-owners.md.Dh2nDSqg.js +0 -13
  410. package/dist/docs/assets/templates_security-reviewer.md.DnwXdnB5.js +0 -3
  411. package/dist/docs/assets/templates_security-reviewer.md.DnwXdnB5.lean.js +0 -1
  412. package/dist/docs/assets/templates_thermo-quality-review.md.MJIdHkzE.js +0 -3
  413. package/dist/docs/assets/templates_thermo-quality-review.md.MJIdHkzE.lean.js +0 -1
  414. package/dist/docs/assets/templates_thermo-review.md.DfN1OFOl.js +0 -3
  415. package/dist/docs/assets/templates_thermo-review.md.DfN1OFOl.lean.js +0 -1
  416. package/dist/docs/assets/templates_triage.md.BYWMloV4.js +0 -9
  417. package/dist/docs/assets/templates_triage.md.BYWMloV4.lean.js +0 -1
  418. package/dist/docs/assets/troubleshooting.md.CGOdNVAL.js +0 -1
  419. package/dist/docs/guides/hooks.html +0 -113
  420. package/dist/docs/guides/hooks.md +0 -116
  421. package/dist/docs/package/skills/create-agent/SKILL.md +0 -178
  422. package/dist/docs/package/skills/evals/SKILL.md +0 -145
  423. package/dist/docs/package/skills/github/SKILL.md +0 -133
  424. package/dist/docs/package/skills/hillclimb/SKILL.md +0 -63
  425. package/dist/docs/package/skills/mcp-auth/SKILL.md +0 -60
  426. package/dist/docs/package/skills/migrate-v2/SKILL.md +0 -394
  427. package/dist/docs/package/skills/setup-slack/SKILL.md +0 -136
  428. package/dist/docs/reference/prompt.html +0 -68
  429. package/dist/docs/reference/prompt.md +0 -47
  430. package/dist/docs/templates/security-reviewer.html +0 -62
  431. package/dist/docs/templates/security-reviewer.md +0 -81
  432. package/dist/docs/templates/thermo-quality-review.html +0 -62
  433. package/dist/docs/templates/thermo-quality-review.md +0 -75
  434. package/dist/docs/templates/thermo-review.html +0 -62
  435. package/dist/docs/templates/thermo-review.md +0 -75
  436. package/dist/docs/templates/triage.html +0 -68
  437. package/dist/docs/templates/triage.md +0 -101
  438. package/dist/internal/filesystem/errors.d.ts +0 -64
  439. package/dist/internal/filesystem/errors.d.ts.map +0 -1
  440. package/dist/internal/filesystem/errors.js +0 -180
  441. package/dist/internal/filesystem/merge.d.ts +0 -34
  442. package/dist/internal/filesystem/merge.d.ts.map +0 -1
  443. package/dist/internal/filesystem/merge.js +0 -154
  444. package/dist/playground/assets/index-Cl9zvmmf.js +0 -62
  445. package/dist/playground/assets/index-F8b5WvWi.css +0 -1
  446. package/docs/guides/hooks.md +0 -121
  447. package/docs/reference/prompt.md +0 -52
  448. package/docs/templates/security-reviewer.md +0 -86
  449. package/docs/templates/thermo-quality-review.md +0 -80
  450. package/docs/templates/thermo-review.md +0 -80
  451. package/docs/templates/triage.md +0 -106
@@ -4,63 +4,61 @@
4
4
  <meta charset="utf-8">
5
5
  <meta name="viewport" content="width=device-width,initial-scale=1">
6
6
  <title>Hillclimbing | Agent SDK</title>
7
- <meta name="description" content="Improve one agent behavior at a time by measuring fixed inputs, changing one lever, and locking each win with an eval.">
7
+ <meta name="description" content="Iteratively improve agent behavior with evals">
8
8
  <meta name="generator" content="VitePress v1.6.4">
9
9
  <link rel="preload stylesheet" href="/docs/assets/style.DK0YX1WL.css" as="style">
10
10
  <link rel="preload stylesheet" href="/docs/vp-icons.css" as="style">
11
11
 
12
- <script type="module" src="/docs/assets/app.DMrZKFjI.js"></script>
12
+ <script type="module" src="/docs/assets/app.BSH__xV4.js"></script>
13
13
  <link rel="preload" href="/docs/assets/inter-roman-latin.Di8DUHzh.woff2" as="font" type="font/woff2" crossorigin="">
14
14
  <link rel="modulepreload" href="/docs/assets/chunks/framework.BNw1pucY.js">
15
- <link rel="modulepreload" href="/docs/assets/chunks/theme.DRciHmS1.js">
15
+ <link rel="modulepreload" href="/docs/assets/chunks/theme.DGq7D4tt.js">
16
16
  <link rel="modulepreload" href="/docs/assets/chunks/katex.DkKDou_j.js">
17
- <link rel="modulepreload" href="/docs/assets/chunks/dagre-KV5264BT.CLMLc0lD.js">
18
- <link rel="modulepreload" href="/docs/assets/chunks/cose-bilkent-S5V4N54A.BfPVKeId.js">
19
- <link rel="modulepreload" href="/docs/assets/chunks/c4Diagram-AHTNJAMY.C1QTb9GM.js">
20
- <link rel="modulepreload" href="/docs/assets/chunks/flowDiagram-DWJPFMVM.MFihq9AF.js">
21
- <link rel="modulepreload" href="/docs/assets/chunks/erDiagram-SMLLAGMA.CsL5JuCV.js">
22
- <link rel="modulepreload" href="/docs/assets/chunks/gitGraphDiagram-UUTBAWPF.1HhlE45Z.js">
23
- <link rel="modulepreload" href="/docs/assets/chunks/ganttDiagram-T4ZO3ILL.CJeLodyG.js">
24
- <link rel="modulepreload" href="/docs/assets/chunks/infoDiagram-42DDH7IO.fp5gvytQ.js">
25
- <link rel="modulepreload" href="/docs/assets/chunks/pieDiagram-DEJITSTG.Ba7yxiW2.js">
26
- <link rel="modulepreload" href="/docs/assets/chunks/quadrantDiagram-34T5L4WZ.Bvskad4p.js">
27
- <link rel="modulepreload" href="/docs/assets/chunks/xychartDiagram-5P7HB3ND.DEH4BcVJ.js">
28
- <link rel="modulepreload" href="/docs/assets/chunks/requirementDiagram-MS252O5E.Ce70ugV9.js">
29
- <link rel="modulepreload" href="/docs/assets/chunks/sequenceDiagram-FGHM5R23.C8a1hDs6.js">
30
- <link rel="modulepreload" href="/docs/assets/chunks/classDiagram-6PBFFD2Q.eClS1Q0E.js">
31
- <link rel="modulepreload" href="/docs/assets/chunks/classDiagram-v2-HSJHXN6E.eClS1Q0E.js">
32
- <link rel="modulepreload" href="/docs/assets/chunks/stateDiagram-FHFEXIEX.D2D8A5CK.js">
33
- <link rel="modulepreload" href="/docs/assets/chunks/stateDiagram-v2-QKLJ7IA2.77yf4kRr.js">
34
- <link rel="modulepreload" href="/docs/assets/chunks/journeyDiagram-VCZTEJTY.qJt4uJ_H.js">
35
- <link rel="modulepreload" href="/docs/assets/chunks/timeline-definition-GMOUNBTQ.C4ePrY8-.js">
36
- <link rel="modulepreload" href="/docs/assets/chunks/mindmap-definition-QFDTVHPH.DgrQRFZW.js">
37
- <link rel="modulepreload" href="/docs/assets/chunks/kanban-definition-6JOO6SKY.z2tG9kly.js">
38
- <link rel="modulepreload" href="/docs/assets/chunks/sankeyDiagram-XADWPNL6.BJ5OEULO.js">
39
- <link rel="modulepreload" href="/docs/assets/chunks/diagram-TYMM5635.BcpVgA_M.js">
40
- <link rel="modulepreload" href="/docs/assets/chunks/diagram-MMDJMWI5.DLnM9eLp.js">
41
- <link rel="modulepreload" href="/docs/assets/chunks/blockDiagram-DXYQGD6D.HjJ2goMX.js">
42
- <link rel="modulepreload" href="/docs/assets/chunks/diagram-5BDNPKRD.C7OF7UfH.js">
43
- <link rel="modulepreload" href="/docs/assets/chunks/architectureDiagram-Q4EWVU46.BcgfYNO_.js">
44
- <link rel="modulepreload" href="/docs/assets/chunks/ishikawaDiagram-UXIWVN3A.C6Wk1V4I.js">
45
- <link rel="modulepreload" href="/docs/assets/chunks/vennDiagram-DHZGUBPP.Dzso6dEA.js">
46
- <link rel="modulepreload" href="/docs/assets/chunks/diagram-G4DWMVQ6.Dpao0Bl1.js">
47
- <link rel="modulepreload" href="/docs/assets/chunks/wardleyDiagram-NUSXRM2D.Bf2ICxjh.js">
17
+ <link rel="modulepreload" href="/docs/assets/chunks/dagre-KV5264BT.Dq1daUru.js">
18
+ <link rel="modulepreload" href="/docs/assets/chunks/cose-bilkent-S5V4N54A.D5za6aH0.js">
19
+ <link rel="modulepreload" href="/docs/assets/chunks/c4Diagram-AHTNJAMY.CyLg6Ovc.js">
20
+ <link rel="modulepreload" href="/docs/assets/chunks/flowDiagram-DWJPFMVM.C58Z9uPK.js">
21
+ <link rel="modulepreload" href="/docs/assets/chunks/erDiagram-SMLLAGMA.DI4TNkye.js">
22
+ <link rel="modulepreload" href="/docs/assets/chunks/gitGraphDiagram-UUTBAWPF.B955oaFl.js">
23
+ <link rel="modulepreload" href="/docs/assets/chunks/ganttDiagram-T4ZO3ILL.DXSpW809.js">
24
+ <link rel="modulepreload" href="/docs/assets/chunks/infoDiagram-42DDH7IO.Dozj_-Gm.js">
25
+ <link rel="modulepreload" href="/docs/assets/chunks/pieDiagram-DEJITSTG.BVYvElmY.js">
26
+ <link rel="modulepreload" href="/docs/assets/chunks/quadrantDiagram-34T5L4WZ.C6Gc2mdY.js">
27
+ <link rel="modulepreload" href="/docs/assets/chunks/xychartDiagram-5P7HB3ND.haEiSxqc.js">
28
+ <link rel="modulepreload" href="/docs/assets/chunks/requirementDiagram-MS252O5E.DWa8b4cn.js">
29
+ <link rel="modulepreload" href="/docs/assets/chunks/sequenceDiagram-FGHM5R23.DlU3vVN9.js">
30
+ <link rel="modulepreload" href="/docs/assets/chunks/classDiagram-6PBFFD2Q.CsJJyx6-.js">
31
+ <link rel="modulepreload" href="/docs/assets/chunks/classDiagram-v2-HSJHXN6E.CsJJyx6-.js">
32
+ <link rel="modulepreload" href="/docs/assets/chunks/stateDiagram-FHFEXIEX.CeqJRqnU.js">
33
+ <link rel="modulepreload" href="/docs/assets/chunks/stateDiagram-v2-QKLJ7IA2.DE0BnUZh.js">
34
+ <link rel="modulepreload" href="/docs/assets/chunks/journeyDiagram-VCZTEJTY.DrqrMhCk.js">
35
+ <link rel="modulepreload" href="/docs/assets/chunks/timeline-definition-GMOUNBTQ.Cqf2kjq9.js">
36
+ <link rel="modulepreload" href="/docs/assets/chunks/mindmap-definition-QFDTVHPH.BBfpYTea.js">
37
+ <link rel="modulepreload" href="/docs/assets/chunks/kanban-definition-6JOO6SKY.Dv-LTrlw.js">
38
+ <link rel="modulepreload" href="/docs/assets/chunks/sankeyDiagram-XADWPNL6.Bi-g4ajv.js">
39
+ <link rel="modulepreload" href="/docs/assets/chunks/diagram-TYMM5635.Dx2FVhCr.js">
40
+ <link rel="modulepreload" href="/docs/assets/chunks/diagram-MMDJMWI5.BOzMEkAr.js">
41
+ <link rel="modulepreload" href="/docs/assets/chunks/blockDiagram-DXYQGD6D.Bs6XxFpe.js">
42
+ <link rel="modulepreload" href="/docs/assets/chunks/diagram-5BDNPKRD.j1YL5BKS.js">
43
+ <link rel="modulepreload" href="/docs/assets/chunks/architectureDiagram-Q4EWVU46.SlHEknkX.js">
44
+ <link rel="modulepreload" href="/docs/assets/chunks/ishikawaDiagram-UXIWVN3A.D2140OzH.js">
45
+ <link rel="modulepreload" href="/docs/assets/chunks/vennDiagram-DHZGUBPP.BeOpE_lM.js">
46
+ <link rel="modulepreload" href="/docs/assets/chunks/diagram-G4DWMVQ6.Bi1aLdhC.js">
47
+ <link rel="modulepreload" href="/docs/assets/chunks/wardleyDiagram-NUSXRM2D.BXS65dAC.js">
48
48
  <link rel="modulepreload" href="/docs/assets/chunks/virtual_mermaid-config.DDnGl6nM.js">
49
- <link rel="modulepreload" href="/docs/assets/hillclimbing.md.By61f3uq.lean.js">
49
+ <link rel="modulepreload" href="/docs/assets/hillclimbing.md.M51bF-oT.lean.js">
50
50
  <script id="check-dark-mode">(()=>{const e=localStorage.getItem("vitepress-theme-appearance")||"auto",a=window.matchMedia("(prefers-color-scheme: dark)").matches;(!e||e==="auto"?a:e==="dark")&&document.documentElement.classList.add("dark")})();</script>
51
51
  <script id="check-mac-os">document.documentElement.classList.toggle("mac",/Mac|iPhone|iPod|iPad/i.test(navigator.platform));</script>
52
52
  <link rel="alternate" type="text/plain" href="/docs/llms.txt">
53
53
  <link rel="alternate" type="text/markdown" href="/docs/hillclimbing.md">
54
54
  </head>
55
55
  <body>
56
- <div id="app"><div class="Layout" data-v-738a160a><!--[--><!--]--><!--[--><span tabindex="-1" data-v-a98c2453></span><a href="#VPContent" class="VPSkipLink visually-hidden" data-v-a98c2453>Skip to content</a><!--]--><!----><header class="VPNav" data-v-738a160a data-v-76ad4059><div class="VPNavBar" data-v-76ad4059 data-v-6fc2afcb><div class="wrapper" data-v-6fc2afcb><div class="container" data-v-6fc2afcb><div class="title" data-v-6fc2afcb><div class="VPNavBarTitle has-sidebar" data-v-6fc2afcb data-v-9e6caa3f><a class="title" href="/docs/" data-v-9e6caa3f><!--[--><!--]--><!----><span data-v-9e6caa3f>Agent SDK</span><!--[--><!--[--><!--[--><!--[--><span class="bdk-version" title="Agent SDK 0.2.29 (@grok/agent-sdk package)" data-v-a03aeddf>0.2.29</span><!--]--><!--]--><!--]--><!--]--></a></div></div><div class="content" data-v-6fc2afcb><div class="content-body" data-v-6fc2afcb><!--[--><!--]--><div class="VPNavBarSearch search" data-v-6fc2afcb><!--[--><!----><div id="local-search"><button type="button" class="DocSearch DocSearch-Button" aria-label="Search"><span class="DocSearch-Button-Container"><span class="vp-icon DocSearch-Search-Icon"></span><span class="DocSearch-Button-Placeholder">Search</span></span><span class="DocSearch-Button-Keys"><kbd class="DocSearch-Button-Key"></kbd><kbd class="DocSearch-Button-Key">K</kbd></span></button></div><!--]--></div><nav aria-labelledby="main-nav-aria-label" class="VPNavBarMenu menu" data-v-6fc2afcb data-v-78f72f99><span id="main-nav-aria-label" class="visually-hidden" data-v-78f72f99> Main Navigation </span><!--[--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/quickstart.html" tabindex="0" data-v-78f72f99 data-v-a7790978><!--[--><span data-v-a7790978>Quickstart</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/guides/mcp-oauth.html" tabindex="0" data-v-78f72f99 data-v-a7790978><!--[--><span data-v-a7790978>Guides</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/guides/github.html" tabindex="0" data-v-78f72f99 data-v-a7790978><!--[--><span data-v-a7790978>Integrations</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/templates/security-reviewer.html" tabindex="0" data-v-78f72f99 data-v-a7790978><!--[--><span data-v-a7790978>Templates</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/reference/project-layout.html" tabindex="0" data-v-78f72f99 data-v-a7790978><!--[--><span data-v-a7790978>Reference</span><!--]--></a><!--]--><!--]--></nav><!----><div class="VPNavBarAppearance appearance" data-v-6fc2afcb data-v-bf7c3d0f><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-bf7c3d0f data-v-e9e166e4 data-v-74f7b0ff><span class="check" data-v-74f7b0ff><span class="icon" data-v-74f7b0ff><!--[--><span class="vpi-sun sun" data-v-e9e166e4></span><span class="vpi-moon moon" data-v-e9e166e4></span><!--]--></span></span></button></div><!----><div class="VPFlyout VPNavBarExtra extra" data-v-6fc2afcb data-v-799c9281 data-v-070c9767><button type="button" class="button" aria-haspopup="true" aria-expanded="false" aria-label="extra navigation" data-v-070c9767><span class="vpi-more-horizontal icon" data-v-070c9767></span></button><div class="menu" data-v-070c9767><div class="VPMenu" data-v-070c9767 data-v-c05d27ee><!----><!--[--><!--[--><!----><div class="group" data-v-799c9281><div class="item appearance" data-v-799c9281><p class="label" data-v-799c9281>Appearance</p><div class="appearance-action" data-v-799c9281><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-799c9281 data-v-e9e166e4 data-v-74f7b0ff><span class="check" data-v-74f7b0ff><span class="icon" data-v-74f7b0ff><!--[--><span class="vpi-sun sun" data-v-e9e166e4></span><span class="vpi-moon moon" data-v-e9e166e4></span><!--]--></span></span></button></div></div></div><!----><!--]--><!--]--></div></div></div><!--[--><!--]--><button type="button" class="VPNavBarHamburger hamburger" aria-label="mobile navigation" aria-expanded="false" aria-controls="VPNavScreen" data-v-6fc2afcb data-v-2caaf84e><span class="container" data-v-2caaf84e><span class="top" data-v-2caaf84e></span><span class="middle" data-v-2caaf84e></span><span class="bottom" data-v-2caaf84e></span></span></button></div></div></div></div><div class="divider" data-v-6fc2afcb><div class="divider-line" data-v-6fc2afcb></div></div></div><!----></header><div class="VPLocalNav has-sidebar empty" data-v-738a160a data-v-e0cbfd80><div class="container" data-v-e0cbfd80><button class="menu" aria-expanded="false" aria-controls="VPSidebarNav" data-v-e0cbfd80><span class="vpi-align-left menu-icon" data-v-e0cbfd80></span><span class="menu-text" data-v-e0cbfd80>Menu</span></button><div class="VPLocalNavOutlineDropdown" style="--vp-vh:0px;" data-v-e0cbfd80 data-v-36e0b8e9><button data-v-36e0b8e9>Return to top</button><!----></div></div></div><aside class="VPSidebar" data-v-738a160a data-v-2859715e><div class="curtain" data-v-2859715e></div><nav class="nav" id="VPSidebarNav" aria-labelledby="sidebar-aria-label" tabindex="-1" data-v-2859715e><span class="visually-hidden" id="sidebar-aria-label" data-v-2859715e> Sidebar Navigation </span><!--[--><!--]--><!--[--><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Getting started</h2><!----></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Overview</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/changelog.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Changelog</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/quickstart.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Building your first agent</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0 collapsible" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Integrations</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-84325fb7><span class="vpi-chevron-right caret-icon" data-v-84325fb7></span></div></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/github.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>GitHub</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/slack.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Slack</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/webhooks.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Webhooks</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/convert-automation.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Cursor Automations</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/opentelemetry.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>OpenTelemetry</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/gitlab.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>GitLab</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/bitbucket.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Bitbucket</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/pagerduty.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>PagerDuty</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0 collapsible" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Extensions</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-84325fb7><span class="vpi-chevron-right caret-icon" data-v-84325fb7></span></div></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/extensions.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Overview</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/cloud-agents.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Cloud agents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/improve.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Self-improvement</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/jev.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Jev</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0 collapsible has-active" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Guides</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-84325fb7><span class="vpi-chevron-right caret-icon" data-v-84325fb7></span></div></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/mcp-oauth.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>MCP</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/hooks.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Hooks</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/agent-to-agent.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Peer agents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/deployment.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Deployment</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/migrate.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Early-alpha migration</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/evals.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Evals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/hillclimbing.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Hillclimbing</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/building-with-agents.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Build with a coding agent</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/troubleshooting.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Troubleshooting</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0 collapsible" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Templates</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-84325fb7><span class="vpi-chevron-right caret-icon" data-v-84325fb7></span></div></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/templates/security-reviewer.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Security reviewer</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/templates/thermo-review.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Thermo review</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/templates/thermo-quality-review.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Thermo quality review</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/templates/agentic-owners.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Agentic Approver</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/templates/triage.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Triage agent</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0 collapsible" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Reference</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-84325fb7><span class="vpi-chevron-right caret-icon" data-v-84325fb7></span></div></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/project-layout.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Project layout</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/agent-config.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Agent config</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/instructions.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Instructions</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/tools.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Tools</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/prompt.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>prompt</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/skills.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Skills</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/connections.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>MCP Connections</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/subagents.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Subagents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/channels.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/schedules.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Schedules & reminders</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/hooks.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Hooks</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/memory.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Memory</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/evals.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Evals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/artifacts.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Artifacts</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/sessions.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Sessions & streaming</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/playground.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Playground</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/cli.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>CLI</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/http-api.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>HTTP API</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/access.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Access</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><!--]--><!--[--><!--]--></nav></aside><div class="VPContent has-sidebar" id="VPContent" data-v-738a160a data-v-9f462cdf><div class="VPDoc has-sidebar has-aside" data-v-9f462cdf data-v-995f90b4><!--[--><!--]--><div class="container" data-v-995f90b4><div class="aside" data-v-995f90b4><div class="aside-curtain" data-v-995f90b4></div><div class="aside-container" data-v-995f90b4><div class="aside-content" data-v-995f90b4><div class="VPDocAside" data-v-995f90b4 data-v-10889960><!--[--><!--]--><!--[--><!--]--><nav aria-labelledby="doc-outline-aria-label" class="VPDocAsideOutline" data-v-10889960 data-v-afb61a88><div class="content" data-v-afb61a88><div class="outline-marker" data-v-afb61a88></div><div aria-level="2" class="outline-title" id="doc-outline-aria-label" role="heading" data-v-afb61a88>On this page</div><ul class="VPDocOutlineItem root" data-v-afb61a88 data-v-0587367d><!--[--><!--]--></ul></div></nav><!--[--><!--]--><div class="spacer" data-v-10889960></div><!--[--><!--]--><!----><!--[--><!--]--><!--[--><!--]--></div></div></div></div><div class="content" data-v-995f90b4><div class="content-container" data-v-995f90b4><!--[--><!--]--><main class="main" data-v-995f90b4><div style="position:relative;" class="vp-doc _docs_hillclimbing" data-v-995f90b4><div><h1 id="hillclimbing" tabindex="-1">Hillclimbing <a class="header-anchor" href="#hillclimbing" aria-label="Permalink to &quot;Hillclimbing&quot;">​</a></h1><p>Hillclimbing improves an agent through controlled before-and-after comparisons. Keep the inputs fixed, name one failure, change one lever, and retain the change only when the same run gets better without crossing the freeze line.</p><p>The companion <a href="./../skills/hillclimb/SKILL.html"><code>hillclimb</code> skill</a> can run this loop with you.</p><h2 id="correct-a-wrong-tool-choice" tabindex="-1">Correct a wrong tool choice <a class="header-anchor" href="#correct-a-wrong-tool-choice" aria-label="Permalink to &quot;Correct a wrong tool choice&quot;">​</a></h2><p>Suppose a PR reviewer calls <code>approve_pr</code> before <code>inspect_pr</code>. Pin one PR fixture, record the original tool order, then tighten the instruction that owns the decision. The improved run should inspect first and leave approval untouched.</p><div class="language-text vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">text</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span>agent-sdk run --dir . \</span></span>
57
- <span class="line"><span> --message &quot;Is https://github.com/acme/checkout/pull/42 ready?&quot;</span></span></code></pre></div><p>Lock the result with the intended and tempting-wrong paths:</p><div class="language-ts vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">ts</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">t.</span><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">succeeded</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">();</span></span>
58
- <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">t.</span><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">calledTool</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">(</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;inspect_pr&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">);</span></span>
59
- <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">t.</span><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">notCalledTool</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">(</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;approve_pr&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">);</span></span></code></pre></div><p>Do not change the tool implementation, prompt, and evidence together. If tool choice improves, you should know which edit caused it.</p><h2 id="reduce-wandering-or-latency" tabindex="-1">Reduce wandering or latency <a class="header-anchor" href="#reduce-wandering-or-latency" aria-label="Permalink to &quot;Reduce wandering or latency&quot;">​</a></h2><p>A slow run often repeats reads, searches broad directories, or asks the model to discover evidence the host already has. Compare the trajectory by tool name and wall time, then move deterministic preparation ahead of the turn or trim the evidence shape.</p><div class="language-text vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">text</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span>before: 14 tool calls, 92 seconds, 6 repeated reads</span></span>
60
- <span class="line"><span>after: 5 tool calls, 31 seconds, 0 repeated reads</span></span></code></pre></div><p>Keep the change only if the answer still finds the same known issue. Efficiency without a quality check can make an agent fast by making it incomplete.</p><h2 id="improve-quality-without-changing-the-contract" tabindex="-1">Improve quality without changing the contract <a class="header-anchor" href="#improve-quality-without-changing-the-contract" aria-label="Permalink to &quot;Improve quality without changing the contract&quot;">​</a></h2><p>For answer quality, freeze the public shape before editing: required sections, API fields, side effects, and existing evals. Compare both runs against a known-good answer or a small set of gold findings, then change the instructions or skill that owns the missing judgment.</p><div class="language-text vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">text</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span>fixture: saved checkout PR diff</span></span>
56
+ <div id="app"><div class="Layout" data-v-738a160a><!--[--><!--]--><!--[--><span tabindex="-1" data-v-a98c2453></span><a href="#VPContent" class="VPSkipLink visually-hidden" data-v-a98c2453>Skip to content</a><!--]--><!----><header class="VPNav" data-v-738a160a data-v-76ad4059><div class="VPNavBar" data-v-76ad4059 data-v-6fc2afcb><div class="wrapper" data-v-6fc2afcb><div class="container" data-v-6fc2afcb><div class="title" data-v-6fc2afcb><div class="VPNavBarTitle has-sidebar" data-v-6fc2afcb data-v-9e6caa3f><a class="title" href="/docs/" data-v-9e6caa3f><!--[--><!--]--><!----><span data-v-9e6caa3f>Agent SDK</span><!--[--><!--[--><!--[--><!--[--><span class="bdk-version" title="Agent SDK 0.4.0 (@grok/agent-sdk package)" data-v-a03aeddf>0.4.0</span><!--]--><!--]--><!--]--><!--]--></a></div></div><div class="content" data-v-6fc2afcb><div class="content-body" data-v-6fc2afcb><!--[--><!--]--><div class="VPNavBarSearch search" data-v-6fc2afcb><!--[--><!----><div id="local-search"><button type="button" class="DocSearch DocSearch-Button" aria-label="Search"><span class="DocSearch-Button-Container"><span class="vp-icon DocSearch-Search-Icon"></span><span class="DocSearch-Button-Placeholder">Search</span></span><span class="DocSearch-Button-Keys"><kbd class="DocSearch-Button-Key"></kbd><kbd class="DocSearch-Button-Key">K</kbd></span></button></div><!--]--></div><nav aria-labelledby="main-nav-aria-label" class="VPNavBarMenu menu" data-v-6fc2afcb data-v-78f72f99><span id="main-nav-aria-label" class="visually-hidden" data-v-78f72f99> Main Navigation </span><!--[--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/quickstart.html" tabindex="0" data-v-78f72f99 data-v-a7790978><!--[--><span data-v-a7790978>Quickstart</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/guides/mcp-oauth.html" tabindex="0" data-v-78f72f99 data-v-a7790978><!--[--><span data-v-a7790978>Guides</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/guides/github.html" tabindex="0" data-v-78f72f99 data-v-a7790978><!--[--><span data-v-a7790978>Integrations</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/templates/reviews.html" tabindex="0" data-v-78f72f99 data-v-a7790978><!--[--><span data-v-a7790978>Templates</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/reference/project-layout.html" tabindex="0" data-v-78f72f99 data-v-a7790978><!--[--><span data-v-a7790978>Reference</span><!--]--></a><!--]--><!--]--></nav><!----><div class="VPNavBarAppearance appearance" data-v-6fc2afcb data-v-bf7c3d0f><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-bf7c3d0f data-v-e9e166e4 data-v-74f7b0ff><span class="check" data-v-74f7b0ff><span class="icon" data-v-74f7b0ff><!--[--><span class="vpi-sun sun" data-v-e9e166e4></span><span class="vpi-moon moon" data-v-e9e166e4></span><!--]--></span></span></button></div><!----><div class="VPFlyout VPNavBarExtra extra" data-v-6fc2afcb data-v-799c9281 data-v-070c9767><button type="button" class="button" aria-haspopup="true" aria-expanded="false" aria-label="extra navigation" data-v-070c9767><span class="vpi-more-horizontal icon" data-v-070c9767></span></button><div class="menu" data-v-070c9767><div class="VPMenu" data-v-070c9767 data-v-c05d27ee><!----><!--[--><!--[--><!----><div class="group" data-v-799c9281><div class="item appearance" data-v-799c9281><p class="label" data-v-799c9281>Appearance</p><div class="appearance-action" data-v-799c9281><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-799c9281 data-v-e9e166e4 data-v-74f7b0ff><span class="check" data-v-74f7b0ff><span class="icon" data-v-74f7b0ff><!--[--><span class="vpi-sun sun" data-v-e9e166e4></span><span class="vpi-moon moon" data-v-e9e166e4></span><!--]--></span></span></button></div></div></div><!----><!--]--><!--]--></div></div></div><!--[--><!--]--><button type="button" class="VPNavBarHamburger hamburger" aria-label="mobile navigation" aria-expanded="false" aria-controls="VPNavScreen" data-v-6fc2afcb data-v-2caaf84e><span class="container" data-v-2caaf84e><span class="top" data-v-2caaf84e></span><span class="middle" data-v-2caaf84e></span><span class="bottom" data-v-2caaf84e></span></span></button></div></div></div></div><div class="divider" data-v-6fc2afcb><div class="divider-line" data-v-6fc2afcb></div></div></div><!----></header><div class="VPLocalNav has-sidebar empty" data-v-738a160a data-v-e0cbfd80><div class="container" data-v-e0cbfd80><button class="menu" aria-expanded="false" aria-controls="VPSidebarNav" data-v-e0cbfd80><span class="vpi-align-left menu-icon" data-v-e0cbfd80></span><span class="menu-text" data-v-e0cbfd80>Menu</span></button><div class="VPLocalNavOutlineDropdown" style="--vp-vh:0px;" data-v-e0cbfd80 data-v-36e0b8e9><button data-v-36e0b8e9>Return to top</button><!----></div></div></div><aside class="VPSidebar" data-v-738a160a data-v-2859715e><div class="curtain" data-v-2859715e></div><nav class="nav" id="VPSidebarNav" aria-labelledby="sidebar-aria-label" tabindex="-1" data-v-2859715e><span class="visually-hidden" id="sidebar-aria-label" data-v-2859715e> Sidebar Navigation </span><!--[--><!--]--><!--[--><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Getting started</h2><!----></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Overview</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/changelog.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Changelog</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/quickstart.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Build a PR reviewer</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0 collapsible" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Integrations</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-84325fb7><span class="vpi-chevron-right caret-icon" data-v-84325fb7></span></div></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/github.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>GitHub</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/slack.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Slack</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/webhooks.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Webhooks</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/convert-automation.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Cursor Automations</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/opentelemetry.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>OpenTelemetry</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/gitlab.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>GitLab</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/bitbucket.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Bitbucket</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/pagerduty.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>PagerDuty</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0 collapsible" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Extensions</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-84325fb7><span class="vpi-chevron-right caret-icon" data-v-84325fb7></span></div></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/extensions.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Overview</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/cloud-agents.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Cloud agents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/improve.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Self-improvement</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/jev.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Jev</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0 collapsible has-active" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Guides</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-84325fb7><span class="vpi-chevron-right caret-icon" data-v-84325fb7></span></div></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/mcp-oauth.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>MCP</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/agent-to-agent.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Peer agents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/deployment.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Deployment</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/guides/migrate.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Early-alpha migration</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/evals.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Evals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/hillclimbing.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Hillclimbing</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/building-with-agents.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Build with a coding agent</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/troubleshooting.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Troubleshooting</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0 collapsible" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Templates</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-84325fb7><span class="vpi-chevron-right caret-icon" data-v-84325fb7></span></div></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/templates/reviews.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Review and triage</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/templates/agentic-owners.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Agentic Approver</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-8ed949fd><section class="VPSidebarItem level-0 collapsible" data-v-8ed949fd data-v-84325fb7><div class="item" role="button" tabindex="0" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><h2 class="text" data-v-84325fb7>Reference</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-84325fb7><span class="vpi-chevron-right caret-icon" data-v-84325fb7></span></div></div><div class="items" data-v-84325fb7><!--[--><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/project-layout.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Project layout</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/agent-config.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Agent config</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/instructions.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Instructions</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/tools.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Tools</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/skills.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Skills</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/connections.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>MCP Connections</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/subagents.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Subagents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/channels.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/schedules.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Schedules & reminders</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/hooks.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Hooks</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/memory.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Memory</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/evals.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Evals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/artifacts.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Artifacts</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/sessions.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Sessions & streaming</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/playground.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Playground</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/cli.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>CLI</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/http-api.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>HTTP API</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/access.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Access</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-84325fb7><div class="item" data-v-84325fb7><div class="indicator" data-v-84325fb7></div><a class="VPLink link link" href="/docs/reference/limits.html" data-v-84325fb7><!--[--><p class="text" data-v-84325fb7>Limits</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><!--]--><!--[--><!--]--></nav></aside><div class="VPContent has-sidebar" id="VPContent" data-v-738a160a data-v-9f462cdf><div class="VPDoc has-sidebar has-aside" data-v-9f462cdf data-v-995f90b4><!--[--><!--]--><div class="container" data-v-995f90b4><div class="aside" data-v-995f90b4><div class="aside-curtain" data-v-995f90b4></div><div class="aside-container" data-v-995f90b4><div class="aside-content" data-v-995f90b4><div class="VPDocAside" data-v-995f90b4 data-v-10889960><!--[--><!--]--><!--[--><!--]--><nav aria-labelledby="doc-outline-aria-label" class="VPDocAsideOutline" data-v-10889960 data-v-afb61a88><div class="content" data-v-afb61a88><div class="outline-marker" data-v-afb61a88></div><div aria-level="2" class="outline-title" id="doc-outline-aria-label" role="heading" data-v-afb61a88>On this page</div><ul class="VPDocOutlineItem root" data-v-afb61a88 data-v-0587367d><!--[--><!--]--></ul></div></nav><!--[--><!--]--><div class="spacer" data-v-10889960></div><!--[--><!--]--><!----><!--[--><!--]--><!--[--><!--]--></div></div></div></div><div class="content" data-v-995f90b4><div class="content-container" data-v-995f90b4><!--[--><!--]--><main class="main" data-v-995f90b4><div style="position:relative;" class="vp-doc _docs_hillclimbing" data-v-995f90b4><div><h1 id="hillclimbing" tabindex="-1">Hillclimbing <a class="header-anchor" href="#hillclimbing" aria-label="Permalink to &quot;Hillclimbing&quot;">​</a></h1><p>Hillclimbing improves an agent one measured change at a time. Hold the inputs fixed, change one thing, and keep it only when the same run gets better without moving the behavior you froze.</p><p>A coding agent in your checkout can run this loop by following the <code>agent-sdk-hillclimb</code> skill.</p><h2 id="run-one-measured-round" tabindex="-1">Run one measured round <a class="header-anchor" href="#run-one-measured-round" aria-label="Permalink to &quot;Run one measured round&quot;">​</a></h2><p>First agree on the agent, one to three fixed inputs, the metric to improve, and the behavior that must not change, such as the output shape, side effects, and existing evals.</p><p>Then run one round:</p><ol><li>Run the agent the way a user would, on saved or local inputs, without deploying or posting live.</li><li>Read the trajectory and name the main failure in one sentence.</li><li>Change one lever.</li><li>Rerun the same fixtures.</li><li>Compare correctness, efficiency, and contract fit.</li><li>Keep, revert, or narrow the change.</li></ol><p>Don&#39;t swap a fixture mid-round, or the comparison means nothing.</p><h2 id="choose-the-smallest-lever" tabindex="-1">Choose the smallest lever <a class="header-anchor" href="#choose-the-smallest-lever" aria-label="Permalink to &quot;Choose the smallest lever&quot;">​</a></h2><p>Start with the earliest place that can remove the failure. Prepare evidence on the host before the turn, then change the shape of that evidence, then the instructions or skills, then remove or gate tools that invite wandering, and change the framework only when the agent can&#39;t express the fix.</p><p>Fixing the host or evidence often improves quality and latency together, because the model searches less for ground truth.</p><h2 id="match-the-fix-to-the-failure" tabindex="-1">Match the fix to the failure <a class="header-anchor" href="#match-the-fix-to-the-failure" aria-label="Permalink to &quot;Match the fix to the failure&quot;">​</a></h2><p>When the agent picks the wrong tool, such as a PR reviewer that calls <code>approve_pr</code> before <code>inspect_pr</code>, pin one PR and tighten only the instruction that owns that decision, so you know which edit fixed it.</p><div class="language-ts vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">ts</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">t.</span><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">calledTool</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">(</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;inspect_pr&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">);</span></span>
57
+ <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">t.</span><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">notCalledTool</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">(</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;approve_pr&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">);</span></span></code></pre></div><p>When a run wanders or runs long, it often repeats reads or makes the model find evidence the host already has. Move that preparation ahead of the turn, and keep the change only if the answer still finds the same issue.</p><div class="language-text vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">text</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span>before: 14 tool calls, 92 seconds, 6 repeated reads</span></span>
58
+ <span class="line"><span>after: 5 tool calls, 31 seconds, 0 repeated reads</span></span></code></pre></div><p>When the answer is weak, freeze the public shape, compare both runs against known-good findings, and edit the instruction or skill that owns the missing judgment.</p><div class="language-text vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">text</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span>fixture: saved checkout PR diff</span></span>
61
59
  <span class="line"><span>success: report both gold findings with file paths</span></span>
62
- <span class="line"><span>freeze: findings JSON shape; no review post; existing smoke evals</span></span></code></pre></div><p>A wording change is not a win by itself. Keep it when correctness rises and every frozen behavior remains intact.</p><h2 id="run-one-measured-round" tabindex="-1">Run one measured round <a class="header-anchor" href="#run-one-measured-round" aria-label="Permalink to &quot;Run one measured round&quot;">​</a></h2><p>Agree on four inputs before editing:</p><ol><li><strong>Target:</strong> the agent path or slug.</li><li><strong>Fixtures:</strong> one to three prompts, saved payloads, or pinned PRs.</li><li><strong>Success:</strong> the metric or decision that should improve.</li><li><strong>Freeze line:</strong> behavior that must not move.</li></ol><p>Then run one loop:</p><ol><li>Exercise the agent through the same surface a user would.</li><li>Read the trajectory and name the dominant failure in one sentence.</li><li>Change one lever.</li><li>Rerun the same fixtures.</li><li>Compare correctness, efficiency, and contract fit.</li><li>Keep, revert, or narrow the change.</li></ol><p>Do not replace a fixture mid-round. A moving input makes the comparison meaningless.</p><h2 id="choose-the-smallest-lever" tabindex="-1">Choose the smallest lever <a class="header-anchor" href="#choose-the-smallest-lever" aria-label="Permalink to &quot;Choose the smallest lever&quot;">​</a></h2><p>Start with the earliest place that can remove the failure:</p><ol><li><strong>Host preparation:</strong> fetch deterministic evidence before the turn.</li><li><strong>Evidence shape:</strong> trim, order, or label what the model receives.</li><li><strong>Instructions or skills:</strong> clarify the decision procedure.</li><li><strong>Tool surface:</strong> remove or gate tools that invite wandering.</li><li><strong>Framework:</strong> change it only when the agent cannot express the fix.</li></ol><p>Changing the host or evidence often improves both quality and latency because the model spends less time searching for ground truth.</p><h2 id="lock-every-kept-win" tabindex="-1">Lock every kept win <a class="header-anchor" href="#lock-every-kept-win" aria-label="Permalink to &quot;Lock every kept win&quot;">​</a></h2><p>Each retained change needs an eval that would have failed before it. Use a tool-choice assertion, <code>maxToolCalls</code> bound, output-shape check, or quality score that captures the measured improvement.</p><div class="language-text vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">text</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span>agent-sdk eval --dir . --tag smoke</span></span></code></pre></div><p>Never weaken an existing gate to make the round pass. The eval is the record that the improvement survived after the fixture and source change leave your working memory.</p><h2 id="run-the-loop-with-a-coding-agent" tabindex="-1">Run the loop with a coding agent <a class="header-anchor" href="#run-the-loop-with-a-coding-agent" aria-label="Permalink to &quot;Run the loop with a coding agent&quot;">​</a></h2><p>Ask the coding agent to read <a href="./../skills/hillclimb/SKILL.html"><code>skills/hillclimb/SKILL.md</code></a>, then give it the target, fixtures, success criterion, and freeze line. It will measure the current run, propose one change, remeasure, and ask before starting another round.</p><p>Use saved or local surfaces during a climb. Do not deploy or post to a real external channel unless that side effect is part of an explicitly approved fixture.</p><h2 id="related" tabindex="-1">Related <a class="header-anchor" href="#related" aria-label="Permalink to &quot;Related&quot;">​</a></h2><ul><li><a href="./evals.html">Evals</a>: author the regression check for a kept win</li><li><a href="./reference/evals.html">Evals reference</a>: trajectory assertions and scores</li><li><a href="./building-with-agents.html">Building agents with agents</a>: delegate creation and verification to a coding agent</li><li><a href="./troubleshooting.html">Fix common agent problems</a>: diagnose a run that is broken instead of underperforming</li></ul></div></div></main><footer class="VPDocFooter" data-v-995f90b4 data-v-8edf72db><!--[--><!--]--><!----><nav class="prev-next" aria-labelledby="doc-footer-aria-label" data-v-8edf72db><span class="visually-hidden" id="doc-footer-aria-label" data-v-8edf72db>Pager</span><div class="pager" data-v-8edf72db><a class="VPLink link pager-link prev" href="/docs/evals.html" data-v-8edf72db><!--[--><span class="desc" data-v-8edf72db>Previous page</span><span class="title" data-v-8edf72db>Evals</span><!--]--></a></div><div class="pager" data-v-8edf72db><a class="VPLink link pager-link next" href="/docs/building-with-agents.html" data-v-8edf72db><!--[--><span class="desc" data-v-8edf72db>Next page</span><span class="title" data-v-8edf72db>Build with a coding agent</span><!--]--></a></div></nav></footer><!--[--><!--]--></div></div></div><!--[--><!--]--></div></div><!----><!--[--><!--]--></div></div>
63
- <script>window.__VP_HASH_MAP__=JSON.parse("{\"building-with-agents.md\":\"Bf1yH2tG\",\"changelog.md\":\"DrLFQUlu\",\"deployment.md\":\"KW0Ni_rY\",\"evals.md\":\"DiljU__Z\",\"guides_agent-to-agent.md\":\"D1D5f6Jc\",\"guides_bitbucket.md\":\"DkjkBRU8\",\"guides_cloud-agents.md\":\"Ceyjv0ul\",\"guides_convert-automation.md\":\"BjraCvok\",\"guides_github.md\":\"CgN5sNbW\",\"guides_gitlab.md\":\"DCE0p7_w\",\"guides_hooks.md\":\"BjeEiM7s\",\"guides_improve.md\":\"BFYtrtLE\",\"guides_jev.md\":\"CEqLgjj9\",\"guides_mcp-oauth.md\":\"DrGY92u4\",\"guides_migrate.md\":\"RS9Eu-Xh\",\"guides_opentelemetry.md\":\"CoZ1kNyd\",\"guides_pagerduty.md\":\"mnKXe9U0\",\"guides_slack.md\":\"BwXHURvY\",\"guides_webhooks.md\":\"BJk63ZTs\",\"hillclimbing.md\":\"By61f3uq\",\"index.md\":\"CAruw_rp\",\"quickstart.md\":\"CljDaJBi\",\"reference_access.md\":\"CEtibb4i\",\"reference_agent-config.md\":\"Ck29qYNj\",\"reference_artifacts.md\":\"C-hKR2hp\",\"reference_channels.md\":\"DsJfGSBJ\",\"reference_cli.md\":\"BkdLRbXg\",\"reference_connections.md\":\"BoAotr7_\",\"reference_evals.md\":\"2BXeMiO2\",\"reference_extensions.md\":\"DI5X5x8p\",\"reference_hooks.md\":\"B-bwJuiG\",\"reference_http-api.md\":\"_RWELr3d\",\"reference_instructions.md\":\"yL2hXLnl\",\"reference_limits.md\":\"BCCnvt3o\",\"reference_memory.md\":\"Dd7cyV9V\",\"reference_playground.md\":\"CQiJk9CM\",\"reference_project-layout.md\":\"BamcPDUn\",\"reference_prompt.md\":\"BMjm1Jp3\",\"reference_schedules.md\":\"RC47aFm4\",\"reference_sessions.md\":\"C7KFhWXV\",\"reference_skills.md\":\"Dej2zQxi\",\"reference_subagents.md\":\"CAhvW4yS\",\"reference_tools.md\":\"DQr7Fnhu\",\"templates_agentic-owners.md\":\"Dh2nDSqg\",\"templates_security-reviewer.md\":\"DnwXdnB5\",\"templates_thermo-quality-review.md\":\"MJIdHkzE\",\"templates_thermo-review.md\":\"DfN1OFOl\",\"templates_triage.md\":\"BYWMloV4\",\"troubleshooting.md\":\"CGOdNVAL\"}");window.__VP_SITE_DATA__=JSON.parse("{\"lang\":\"en-US\",\"dir\":\"ltr\",\"title\":\"Agent SDK\",\"description\":\"Build and serve bots and agents from versioned Markdown and TypeScript.\",\"base\":\"/docs/\",\"head\":[],\"router\":{\"prefetchLinks\":true},\"appearance\":true,\"themeConfig\":{\"sdkVersion\":\"0.2.29\",\"nav\":[{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Guides\",\"link\":\"/guides/mcp-oauth\"},{\"text\":\"Integrations\",\"link\":\"/guides/github\"},{\"text\":\"Templates\",\"link\":\"/templates/security-reviewer\"},{\"text\":\"Reference\",\"link\":\"/reference/project-layout\"}],\"search\":{\"provider\":\"local\"},\"outline\":{\"level\":[2,3]},\"sidebar\":[{\"text\":\"Getting started\",\"items\":[{\"text\":\"Overview\",\"link\":\"/\"},{\"text\":\"Changelog\",\"link\":\"/changelog\"},{\"text\":\"Building your first agent\",\"link\":\"/quickstart\"}]},{\"text\":\"Integrations\",\"collapsed\":false,\"items\":[{\"text\":\"GitHub\",\"link\":\"/guides/github\"},{\"text\":\"Slack\",\"link\":\"/guides/slack\"},{\"text\":\"Webhooks\",\"link\":\"/guides/webhooks\"},{\"text\":\"Cursor Automations\",\"link\":\"/guides/convert-automation\"},{\"text\":\"OpenTelemetry\",\"link\":\"/guides/opentelemetry\"},{\"text\":\"GitLab\",\"link\":\"/guides/gitlab\"},{\"text\":\"Bitbucket\",\"link\":\"/guides/bitbucket\"},{\"text\":\"PagerDuty\",\"link\":\"/guides/pagerduty\"}]},{\"text\":\"Extensions\",\"collapsed\":false,\"items\":[{\"text\":\"Overview\",\"link\":\"/reference/extensions\"},{\"text\":\"Cloud agents\",\"link\":\"/guides/cloud-agents\"},{\"text\":\"Self-improvement\",\"link\":\"/guides/improve\"},{\"text\":\"Jev\",\"link\":\"/guides/jev\"}]},{\"text\":\"Guides\",\"collapsed\":false,\"items\":[{\"text\":\"MCP\",\"link\":\"/guides/mcp-oauth\"},{\"text\":\"Hooks\",\"link\":\"/guides/hooks\"},{\"text\":\"Peer agents\",\"link\":\"/guides/agent-to-agent\"},{\"text\":\"Deployment\",\"link\":\"/deployment\"},{\"text\":\"Early-alpha migration\",\"link\":\"/guides/migrate\"},{\"text\":\"Evals\",\"link\":\"/evals\"},{\"text\":\"Hillclimbing\",\"link\":\"/hillclimbing\"},{\"text\":\"Build with a coding agent\",\"link\":\"/building-with-agents\"},{\"text\":\"Troubleshooting\",\"link\":\"/troubleshooting\"}]},{\"text\":\"Templates\",\"collapsed\":false,\"items\":[{\"text\":\"Security reviewer\",\"link\":\"/templates/security-reviewer\"},{\"text\":\"Thermo review\",\"link\":\"/templates/thermo-review\"},{\"text\":\"Thermo quality review\",\"link\":\"/templates/thermo-quality-review\"},{\"text\":\"Agentic Approver\",\"link\":\"/templates/agentic-owners\"},{\"text\":\"Triage agent\",\"link\":\"/templates/triage\"}]},{\"text\":\"Reference\",\"collapsed\":false,\"items\":[{\"text\":\"Project layout\",\"link\":\"/reference/project-layout\"},{\"text\":\"Agent config\",\"link\":\"/reference/agent-config\"},{\"text\":\"Instructions\",\"link\":\"/reference/instructions\"},{\"text\":\"Tools\",\"link\":\"/reference/tools\"},{\"text\":\"prompt\",\"link\":\"/reference/prompt\"},{\"text\":\"Skills\",\"link\":\"/reference/skills\"},{\"text\":\"MCP Connections\",\"link\":\"/reference/connections\"},{\"text\":\"Subagents\",\"link\":\"/reference/subagents\"},{\"text\":\"Channels\",\"link\":\"/reference/channels\"},{\"text\":\"Schedules & reminders\",\"link\":\"/reference/schedules\"},{\"text\":\"Hooks\",\"link\":\"/reference/hooks\"},{\"text\":\"Memory\",\"link\":\"/reference/memory\"},{\"text\":\"Evals\",\"link\":\"/reference/evals\"},{\"text\":\"Artifacts\",\"link\":\"/reference/artifacts\"},{\"text\":\"Sessions & streaming\",\"link\":\"/reference/sessions\"},{\"text\":\"Playground\",\"link\":\"/reference/playground\"},{\"text\":\"CLI\",\"link\":\"/reference/cli\"},{\"text\":\"HTTP API\",\"link\":\"/reference/http-api\"},{\"text\":\"Access\",\"link\":\"/reference/access\"}]}]},\"locales\":{},\"scrollOffset\":134,\"cleanUrls\":false}");</script>
60
+ <span class="line"><span>freeze: findings JSON shape; no review post; existing smoke evals</span></span></code></pre></div><h2 id="lock-every-kept-win" tabindex="-1">Lock every kept win <a class="header-anchor" href="#lock-every-kept-win" aria-label="Permalink to &quot;Lock every kept win&quot;">​</a></h2><p>Each kept change needs an eval that would have failed before it, such as a tool-choice assertion, a <code>maxToolCalls</code> bound, an output-shape check, or a quality score.</p><div class="language-bash vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">bash</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agent-sdk</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> eval</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --dir</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> .</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --tag</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> smoke</span></span></code></pre></div><p>Don&#39;t loosen an existing eval so that a new change passes. If an older eval starts failing, the change broke something that used to work, and you should fix the change instead.</p></div></div></main><footer class="VPDocFooter" data-v-995f90b4 data-v-8edf72db><!--[--><!--]--><!----><nav class="prev-next" aria-labelledby="doc-footer-aria-label" data-v-8edf72db><span class="visually-hidden" id="doc-footer-aria-label" data-v-8edf72db>Pager</span><div class="pager" data-v-8edf72db><a class="VPLink link pager-link prev" href="/docs/evals.html" data-v-8edf72db><!--[--><span class="desc" data-v-8edf72db>Previous page</span><span class="title" data-v-8edf72db>Evals</span><!--]--></a></div><div class="pager" data-v-8edf72db><a class="VPLink link pager-link next" href="/docs/building-with-agents.html" data-v-8edf72db><!--[--><span class="desc" data-v-8edf72db>Next page</span><span class="title" data-v-8edf72db>Build with a coding agent</span><!--]--></a></div></nav></footer><!--[--><!--]--></div></div></div><!--[--><!--]--></div></div><!----><!--[--><!--]--></div></div>
61
+ <script>window.__VP_HASH_MAP__=JSON.parse("{\"building-with-agents.md\":\"Bo3ct-rf\",\"changelog.md\":\"b6H_ea44\",\"deployment.md\":\"DJtuDVLh\",\"evals.md\":\"Dne6VRmZ\",\"guides_agent-to-agent.md\":\"DcB3nhOz\",\"guides_bitbucket.md\":\"BiA6ZMFW\",\"guides_cloud-agents.md\":\"CiXAW6on\",\"guides_convert-automation.md\":\"BvTxsJ8J\",\"guides_github.md\":\"CcuZ7keS\",\"guides_gitlab.md\":\"DcKusOcU\",\"guides_improve.md\":\"o3WOGR7n\",\"guides_jev.md\":\"DGCzkMUN\",\"guides_mcp-oauth.md\":\"GVPMKcYl\",\"guides_migrate.md\":\"B8CZxnnX\",\"guides_opentelemetry.md\":\"BRSktBHt\",\"guides_pagerduty.md\":\"pMXUkZZc\",\"guides_slack.md\":\"D7Rwyva4\",\"guides_webhooks.md\":\"B5DXt2v1\",\"hillclimbing.md\":\"M51bF-oT\",\"index.md\":\"DnJPj_B6\",\"quickstart.md\":\"r-oE8nkX\",\"reference_access.md\":\"C5lblfUC\",\"reference_agent-config.md\":\"BSNp2uJ7\",\"reference_artifacts.md\":\"BJpAI7h9\",\"reference_channels.md\":\"CPWOpdsC\",\"reference_cli.md\":\"DHm6DmYk\",\"reference_connections.md\":\"DoY4vrDH\",\"reference_evals.md\":\"DgfUtfIc\",\"reference_extensions.md\":\"DHbW6eOf\",\"reference_hooks.md\":\"vbIM394-\",\"reference_http-api.md\":\"C7Cb0r2J\",\"reference_instructions.md\":\"CV6Qq0UX\",\"reference_limits.md\":\"CbdtFr4T\",\"reference_memory.md\":\"CdoJISM-\",\"reference_playground.md\":\"DIUsRAcy\",\"reference_project-layout.md\":\"9DiKoE16\",\"reference_schedules.md\":\"Dd8_d0WG\",\"reference_sessions.md\":\"NXVQt9mP\",\"reference_skills.md\":\"DzCPNPkG\",\"reference_subagents.md\":\"Dvpq2sUu\",\"reference_tools.md\":\"B2OaR1Nb\",\"templates_agentic-owners.md\":\"B6ZmKAzT\",\"templates_reviews.md\":\"dr9Xv_KM\",\"troubleshooting.md\":\"DfKKvW1R\"}");window.__VP_SITE_DATA__=JSON.parse("{\"lang\":\"en-US\",\"dir\":\"ltr\",\"title\":\"Agent SDK\",\"description\":\"Build and serve bots and agents from versioned Markdown and TypeScript.\",\"base\":\"/docs/\",\"head\":[],\"router\":{\"prefetchLinks\":true},\"appearance\":true,\"themeConfig\":{\"sdkVersion\":\"0.4.0\",\"nav\":[{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Guides\",\"link\":\"/guides/mcp-oauth\"},{\"text\":\"Integrations\",\"link\":\"/guides/github\"},{\"text\":\"Templates\",\"link\":\"/templates/reviews\"},{\"text\":\"Reference\",\"link\":\"/reference/project-layout\"}],\"search\":{\"provider\":\"local\"},\"outline\":{\"level\":[2,3]},\"sidebar\":[{\"text\":\"Getting started\",\"items\":[{\"text\":\"Overview\",\"link\":\"/\"},{\"text\":\"Changelog\",\"link\":\"/changelog\"},{\"text\":\"Build a PR reviewer\",\"link\":\"/quickstart\"}]},{\"text\":\"Integrations\",\"collapsed\":false,\"items\":[{\"text\":\"GitHub\",\"link\":\"/guides/github\"},{\"text\":\"Slack\",\"link\":\"/guides/slack\"},{\"text\":\"Webhooks\",\"link\":\"/guides/webhooks\"},{\"text\":\"Cursor Automations\",\"link\":\"/guides/convert-automation\"},{\"text\":\"OpenTelemetry\",\"link\":\"/guides/opentelemetry\"},{\"text\":\"GitLab\",\"link\":\"/guides/gitlab\"},{\"text\":\"Bitbucket\",\"link\":\"/guides/bitbucket\"},{\"text\":\"PagerDuty\",\"link\":\"/guides/pagerduty\"}]},{\"text\":\"Extensions\",\"collapsed\":false,\"items\":[{\"text\":\"Overview\",\"link\":\"/reference/extensions\"},{\"text\":\"Cloud agents\",\"link\":\"/guides/cloud-agents\"},{\"text\":\"Self-improvement\",\"link\":\"/guides/improve\"},{\"text\":\"Jev\",\"link\":\"/guides/jev\"}]},{\"text\":\"Guides\",\"collapsed\":false,\"items\":[{\"text\":\"MCP\",\"link\":\"/guides/mcp-oauth\"},{\"text\":\"Peer agents\",\"link\":\"/guides/agent-to-agent\"},{\"text\":\"Deployment\",\"link\":\"/deployment\"},{\"text\":\"Early-alpha migration\",\"link\":\"/guides/migrate\"},{\"text\":\"Evals\",\"link\":\"/evals\"},{\"text\":\"Hillclimbing\",\"link\":\"/hillclimbing\"},{\"text\":\"Build with a coding agent\",\"link\":\"/building-with-agents\"},{\"text\":\"Troubleshooting\",\"link\":\"/troubleshooting\"}]},{\"text\":\"Templates\",\"collapsed\":false,\"items\":[{\"text\":\"Review and triage\",\"link\":\"/templates/reviews\"},{\"text\":\"Agentic Approver\",\"link\":\"/templates/agentic-owners\"}]},{\"text\":\"Reference\",\"collapsed\":false,\"items\":[{\"text\":\"Project layout\",\"link\":\"/reference/project-layout\"},{\"text\":\"Agent config\",\"link\":\"/reference/agent-config\"},{\"text\":\"Instructions\",\"link\":\"/reference/instructions\"},{\"text\":\"Tools\",\"link\":\"/reference/tools\"},{\"text\":\"Skills\",\"link\":\"/reference/skills\"},{\"text\":\"MCP Connections\",\"link\":\"/reference/connections\"},{\"text\":\"Subagents\",\"link\":\"/reference/subagents\"},{\"text\":\"Channels\",\"link\":\"/reference/channels\"},{\"text\":\"Schedules & reminders\",\"link\":\"/reference/schedules\"},{\"text\":\"Hooks\",\"link\":\"/reference/hooks\"},{\"text\":\"Memory\",\"link\":\"/reference/memory\"},{\"text\":\"Evals\",\"link\":\"/reference/evals\"},{\"text\":\"Artifacts\",\"link\":\"/reference/artifacts\"},{\"text\":\"Sessions & streaming\",\"link\":\"/reference/sessions\"},{\"text\":\"Playground\",\"link\":\"/reference/playground\"},{\"text\":\"CLI\",\"link\":\"/reference/cli\"},{\"text\":\"HTTP API\",\"link\":\"/reference/http-api\"},{\"text\":\"Access\",\"link\":\"/reference/access\"},{\"text\":\"Limits\",\"link\":\"/reference/limits\"}]}]},\"locales\":{},\"scrollOffset\":134,\"cleanUrls\":false}");</script>
64
62
 
65
63
  </body>
66
64
  </html>
@@ -1,58 +1,57 @@
1
1
  # Hillclimbing
2
2
 
3
- Hillclimbing improves an agent through controlled before-and-after
4
- comparisons. Keep the inputs fixed, name one failure, change one lever,
5
- and retain the change only when the same run gets better without
6
- crossing the freeze line.
3
+ Hillclimbing improves an agent one measured change at a time. Hold the inputs fixed, change one
4
+ thing, and keep it only when the same run gets better without moving the behavior you froze.
7
5
 
8
- The companion [`hillclimb` skill](/docs/package/skills/hillclimb/SKILL.md) can run
9
- this loop with you.
6
+ A coding agent in your checkout can run this loop by following the `agent-sdk-hillclimb` skill.
10
7
 
11
- ## Correct a wrong tool choice
8
+ ## Run one measured round
12
9
 
13
- Suppose a PR reviewer calls `approve_pr` before `inspect_pr`. Pin one PR
14
- fixture, record the original tool order, then tighten the instruction
15
- that owns the decision. The improved run should inspect first and leave
16
- approval untouched.
10
+ First agree on the agent, one to three fixed inputs, the metric to improve, and the behavior that
11
+ must not change, such as the output shape, side effects, and existing evals.
17
12
 
18
- ```text
19
- agent-sdk run --dir . \
20
- --message "Is https://github.com/acme/checkout/pull/42 ready?"
21
- ```
13
+ Then run one round:
14
+
15
+ 1. Run the agent the way a user would, on saved or local inputs, without deploying or posting live.
16
+ 2. Read the trajectory and name the main failure in one sentence.
17
+ 3. Change one lever.
18
+ 4. Rerun the same fixtures.
19
+ 5. Compare correctness, efficiency, and contract fit.
20
+ 6. Keep, revert, or narrow the change.
21
+
22
+ Don't swap a fixture mid-round, or the comparison means nothing.
23
+
24
+ ## Choose the smallest lever
22
25
 
23
- Lock the result with the intended and tempting-wrong paths:
26
+ Start with the earliest place that can remove the failure. Prepare evidence on the host before the
27
+ turn, then change the shape of that evidence, then the instructions or skills, then remove or gate
28
+ tools that invite wandering, and change the framework only when the agent can't express the fix.
29
+
30
+ Fixing the host or evidence often improves quality and latency together, because the model searches
31
+ less for ground truth.
32
+
33
+ ## Match the fix to the failure
34
+
35
+ When the agent picks the wrong tool, such as a PR reviewer that calls `approve_pr` before
36
+ `inspect_pr`, pin one PR and tighten only the instruction that owns that decision, so you know which
37
+ edit fixed it.
24
38
 
25
39
  ```ts
26
- t.succeeded();
27
40
  t.calledTool("inspect_pr");
28
41
  t.notCalledTool("approve_pr");
29
42
  ```
30
43
 
31
- Do not change the tool implementation, prompt, and evidence together.
32
- If tool choice improves, you should know which edit caused it.
33
-
34
- ## Reduce wandering or latency
35
-
36
- A slow run often repeats reads, searches broad directories, or asks the
37
- model to discover evidence the host already has. Compare the trajectory
38
- by tool name and wall time, then move deterministic preparation ahead
39
- of the turn or trim the evidence shape.
44
+ When a run wanders or runs long, it often repeats reads or makes the model find evidence the host
45
+ already has. Move that preparation ahead of the turn, and keep the change only if the answer still
46
+ finds the same issue.
40
47
 
41
48
  ```text
42
49
  before: 14 tool calls, 92 seconds, 6 repeated reads
43
50
  after: 5 tool calls, 31 seconds, 0 repeated reads
44
51
  ```
45
52
 
46
- Keep the change only if the answer still finds the same known issue.
47
- Efficiency without a quality check can make an agent fast by making it
48
- incomplete.
49
-
50
- ## Improve quality without changing the contract
51
-
52
- For answer quality, freeze the public shape before editing: required
53
- sections, API fields, side effects, and existing evals. Compare both
54
- runs against a known-good answer or a small set of gold findings, then
55
- change the instructions or skill that owns the missing judgment.
53
+ When the answer is weak, freeze the public shape, compare both runs against known-good findings, and
54
+ edit the instruction or skill that owns the missing judgment.
56
55
 
57
56
  ```text
58
57
  fixture: saved checkout PR diff
@@ -60,75 +59,14 @@ success: report both gold findings with file paths
60
59
  freeze: findings JSON shape; no review post; existing smoke evals
61
60
  ```
62
61
 
63
- A wording change is not a win by itself. Keep it when correctness rises
64
- and every frozen behavior remains intact.
65
-
66
- ## Run one measured round
67
-
68
- Agree on four inputs before editing:
69
-
70
- 1. **Target:** the agent path or slug.
71
- 2. **Fixtures:** one to three prompts, saved payloads, or pinned PRs.
72
- 3. **Success:** the metric or decision that should improve.
73
- 4. **Freeze line:** behavior that must not move.
74
-
75
- Then run one loop:
76
-
77
- 1. Exercise the agent through the same surface a user would.
78
- 2. Read the trajectory and name the dominant failure in one sentence.
79
- 3. Change one lever.
80
- 4. Rerun the same fixtures.
81
- 5. Compare correctness, efficiency, and contract fit.
82
- 6. Keep, revert, or narrow the change.
83
-
84
- Do not replace a fixture mid-round. A moving input makes the comparison
85
- meaningless.
86
-
87
- ## Choose the smallest lever
88
-
89
- Start with the earliest place that can remove the failure:
90
-
91
- 1. **Host preparation:** fetch deterministic evidence before the turn.
92
- 2. **Evidence shape:** trim, order, or label what the model receives.
93
- 3. **Instructions or skills:** clarify the decision procedure.
94
- 4. **Tool surface:** remove or gate tools that invite wandering.
95
- 5. **Framework:** change it only when the agent cannot express the fix.
96
-
97
- Changing the host or evidence often improves both quality and latency
98
- because the model spends less time searching for ground truth.
99
-
100
62
  ## Lock every kept win
101
63
 
102
- Each retained change needs an eval that would have failed before it. Use
103
- a tool-choice assertion, `maxToolCalls` bound, output-shape check, or
104
- quality score that captures the measured improvement.
64
+ Each kept change needs an eval that would have failed before it, such as a tool-choice assertion, a
65
+ `maxToolCalls` bound, an output-shape check, or a quality score.
105
66
 
106
- ```text
67
+ ```bash
107
68
  agent-sdk eval --dir . --tag smoke
108
69
  ```
109
70
 
110
- Never weaken an existing gate to make the round pass. The eval is the
111
- record that the improvement survived after the fixture and source
112
- change leave your working memory.
113
-
114
- ## Run the loop with a coding agent
115
-
116
- Ask the coding agent to read
117
- [`skills/hillclimb/SKILL.md`](/docs/package/skills/hillclimb/SKILL.md), then give
118
- it the target, fixtures, success criterion, and freeze line. It will
119
- measure the current run, propose one change, remeasure, and ask before
120
- starting another round.
121
-
122
- Use saved or local surfaces during a climb. Do not deploy or post to a
123
- real external channel unless that side effect is part of an explicitly
124
- approved fixture.
125
-
126
- ## Related
127
-
128
- - [Evals](/docs/evals.md): author the regression check for a kept win
129
- - [Evals reference](/docs/reference/evals.md): trajectory assertions and
130
- scores
131
- - [Building agents with agents](/docs/building-with-agents.md): delegate
132
- creation and verification to a coding agent
133
- - [Fix common agent problems](/docs/troubleshooting.md): diagnose a run
134
- that is broken instead of underperforming
71
+ Don't loosen an existing eval so that a new change passes. If an older eval starts failing, the
72
+ change broke something that used to work, and you should fix the change instead.