crysk 0.6.1 → 0.8.0

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 (179) hide show
  1. package/README.md +28 -22
  2. package/dist/SKILL-1q3658n0.md +74 -0
  3. package/dist/SKILL-25j4q9b8.md +143 -0
  4. package/dist/SKILL-32cmme3j.md +143 -0
  5. package/dist/SKILL-4747bhne.md +143 -0
  6. package/dist/SKILL-4jc3sv23.md +74 -0
  7. package/dist/SKILL-c0492rch.md +540 -0
  8. package/dist/SKILL-d1nzkhhr.md +74 -0
  9. package/dist/SKILL-f8p7kvx0.md +143 -0
  10. package/dist/SKILL-gxd45g38.md +553 -0
  11. package/dist/SKILL-rec41991.md +66 -0
  12. package/dist/SKILL-zze8zj5g.md +536 -0
  13. package/dist/adjudicator-eckqh14e.md +78 -0
  14. package/dist/ci-admission-qn797a6c.md +128 -0
  15. package/dist/cli.js +18608 -18523
  16. package/dist/consolidate-4xw1pkg7.md +24 -0
  17. package/dist/deep-review-logic-dgr112y1.md +196 -0
  18. package/dist/deep-review-logic-fck5z4xe.md +197 -0
  19. package/dist/deep-review-logic-h1fpamfh.md +192 -0
  20. package/dist/deep-review-logic-m4vk5683.md +196 -0
  21. package/dist/deep-review-logic-z63kmgmc.md +157 -0
  22. package/dist/deep-review-parity-7x7s5d81.md +148 -0
  23. package/dist/deep-review-parity-9wxmynsc.md +115 -0
  24. package/dist/deep-review-parity-bwwe7cx6.md +152 -0
  25. package/dist/deep-review-parity-e50pw3pk.md +117 -0
  26. package/dist/deep-review-parity-ennmm9m1.md +151 -0
  27. package/{prompts/default/deep-review-parity.md → dist/deep-review-parity-h7y6y2he.md} +14 -16
  28. package/dist/deep-review-parity-xg6wy69c.md +151 -0
  29. package/dist/deep-review-reliability-6h3c4gmw.md +125 -0
  30. package/dist/deep-review-reliability-dcjab69x.md +122 -0
  31. package/dist/deep-review-reliability-ec7b5pbk.md +159 -0
  32. package/dist/deep-review-reliability-hvasvmt9.md +153 -0
  33. package/dist/deep-review-reliability-mjnayxmt.md +158 -0
  34. package/dist/deep-review-reliability-mm7fnayf.md +124 -0
  35. package/dist/deep-review-reliability-xk4xvc2w.md +158 -0
  36. package/dist/deep-review-resilience-0esd5qeh.md +148 -0
  37. package/dist/deep-review-resilience-envgcmpy.md +115 -0
  38. package/dist/deep-review-resilience-fgx8ythf.md +151 -0
  39. package/dist/deep-review-resilience-mq5qa6ad.md +152 -0
  40. package/dist/deep-review-resilience-pwh1gawp.md +151 -0
  41. package/dist/deep-review-resilience-s9n21dkx.md +117 -0
  42. package/dist/deep-review-resilience-vz4qskea.md +119 -0
  43. package/dist/deep-review-security-1vwjdc9q.md +205 -0
  44. package/dist/deep-review-security-x1cc1z00.md +205 -0
  45. package/dist/index.js +590 -485
  46. package/dist/opencode-ci-jc8mf4wn.md +106 -0
  47. package/dist/opencode-ci-qrcv7a99.md +106 -0
  48. package/dist/opencode-ci-rdpcc7xc.md +106 -0
  49. package/dist/opencode-ci-yt0k74t0.md +106 -0
  50. package/dist/review-refuter-7qffwtzz.md +77 -0
  51. package/dist/verifier-tyh8r45v.md +40 -0
  52. package/dist/verifier-xf4xj701.md +40 -0
  53. package/dist/workflow-andv2dvm.yml +158 -0
  54. package/package.json +1 -1
  55. package/prompts/consolidate.md +1 -1
  56. package/prompts/default/PROVENANCE.md +6 -0
  57. package/prompts/default/deep-review-logic.md +56 -15
  58. package/prompts/default/deep-review-reliability.md +64 -30
  59. package/prompts/default/deep-review-resilience.md +71 -37
  60. package/prompts/default/deep-review-security.md +205 -0
  61. package/prompts/default/review-refuter.md +1 -1
  62. package/prompts/verifier.md +40 -0
  63. package/skills/crysk/SKILL.md +6 -6
  64. package/skills/crysk-ci-setup/SKILL.md +32 -33
  65. package/skills/crysk-ci-setup/assets/workflow.yml +5 -10
  66. package/skills/crysk-ci-setup/references/ci-admission.md +5 -5
  67. package/skills/crysk-ci-setup/references/opencode-ci.md +6 -6
  68. package/skills/crysk-loop/SKILL.md +37 -54
  69. package/skills/crysk-loop/adjudicator.md +10 -18
  70. package/skills/crysk-pr/SKILL.md +1 -1
  71. package/src/asset-manifest.ts +6 -6
  72. package/src/assets.ts +4 -4
  73. package/src/brand.ts +20 -30
  74. package/src/ci/gates.ts +155 -192
  75. package/src/ci/outcome.ts +4 -31
  76. package/src/ci/platform-limits.ts +135 -0
  77. package/src/ci/platform-report.ts +212 -126
  78. package/src/ci/platform.ts +28 -26
  79. package/src/ci/reporter.ts +25 -26
  80. package/src/ci/review-admission.ts +1 -2
  81. package/src/ci/review-levels.ts +12 -53
  82. package/src/ci/setup.ts +96 -66
  83. package/src/ci/trace-bundle.ts +11 -27
  84. package/src/ci/trace-transcripts.ts +4 -6
  85. package/src/ci/trigger.ts +3 -5
  86. package/src/cli.ts +6 -5
  87. package/src/commands/ci-setup-questions.ts +2 -55
  88. package/src/commands/ci-setup.ts +2 -26
  89. package/src/commands/triage.ts +131 -38
  90. package/src/commands/usage.ts +0 -1
  91. package/src/compare/compare.ts +3 -3
  92. package/src/compare/floor-test.ts +30 -201
  93. package/src/compare/hunter-exam.ts +631 -0
  94. package/src/config/config.ts +6 -30
  95. package/src/diversity/accounting.ts +0 -26
  96. package/src/diversity/benchmark.ts +1 -17
  97. package/src/diversity/identity.ts +0 -14
  98. package/src/diversity/pipeline-integration.ts +0 -29
  99. package/src/doctor.ts +3 -6
  100. package/src/execution/claude-session-trace.ts +7 -2
  101. package/src/execution/contracts.ts +15 -9
  102. package/src/execution/harness.ts +165 -187
  103. package/src/execution/spend-limiter.ts +39 -131
  104. package/src/execution/usage-facts.ts +397 -0
  105. package/src/execution/usage-normalized.ts +33 -86
  106. package/src/harness/materiality-judge.ts +0 -5
  107. package/src/harness/pipeline-plan.ts +0 -3
  108. package/src/harness/replay-rows.ts +0 -5
  109. package/src/harness/route-runner.ts +2 -2
  110. package/src/harness/single-step.ts +3 -9
  111. package/src/index.ts +1 -1
  112. package/src/mcp/preflight.ts +2 -2
  113. package/src/model/catalog.ts +1 -3
  114. package/src/model/free-discovery.ts +11 -78
  115. package/src/model/model-name.ts +16 -5
  116. package/src/model/profiles/claude.ts +1 -2
  117. package/src/model/profiles/openai.ts +1 -2
  118. package/src/model/profiles.ts +2 -3
  119. package/src/model/provider-capabilities.ts +16 -55
  120. package/src/model/routing.ts +39 -36
  121. package/src/pr/admission.ts +6 -6
  122. package/src/pr/answered-findings.ts +8 -4
  123. package/src/pr/ci-publish.ts +0 -2
  124. package/src/pr/discovery.ts +4 -25
  125. package/src/pr/plan.ts +40 -121
  126. package/src/pr/pr-status.ts +14 -2
  127. package/src/pr/pr.ts +1 -1
  128. package/src/pr/preflight.ts +44 -15
  129. package/src/pr/publish-outcome.ts +1 -19
  130. package/src/pr/review-pr.ts +15 -47
  131. package/src/pr/target.ts +76 -82
  132. package/src/production-runtime.ts +52 -114
  133. package/src/progress.ts +0 -54
  134. package/src/rereview/prepare.ts +4 -1
  135. package/src/rereview/state.ts +1 -1
  136. package/src/rereview/verify.ts +15 -9
  137. package/src/review/boundary.ts +5 -8
  138. package/src/review/drafts.ts +3 -3
  139. package/src/review/findings-conformance.ts +1 -1
  140. package/src/review/findings.ts +0 -4
  141. package/src/review/pipeline.ts +194 -407
  142. package/src/review/preflight.ts +175 -201
  143. package/src/review/prompt-set.ts +1 -2
  144. package/src/review/report.ts +2 -162
  145. package/src/review/review.ts +25 -74
  146. package/src/review/route-preflight.ts +19 -35
  147. package/src/review/run.ts +37 -73
  148. package/src/review/size-gate.ts +3 -3
  149. package/src/review/spec.ts +6 -11
  150. package/src/review/step-runner.ts +4 -4
  151. package/src/runner-authority.ts +10 -9
  152. package/src/security/credential-broker.ts +5 -6
  153. package/src/server/preflight.ts +0 -1
  154. package/src/server/server.ts +17 -6
  155. package/src/store/activity.ts +1 -2
  156. package/src/store/backfill-preflight.ts +0 -3
  157. package/src/store/backfill.ts +0 -1
  158. package/src/store/metrics-preflight.ts +4 -14
  159. package/src/store/metrics.ts +5 -5
  160. package/src/store/preflight.ts +0 -13
  161. package/src/store/store.ts +11 -12
  162. package/src/transport-registry.ts +6 -30
  163. package/src/transports/claude-code-cli.ts +149 -215
  164. package/src/transports/opencode-client.ts +4 -24
  165. package/src/transports/opencode-sdk.ts +17 -112
  166. package/src/triage/plain-reply.ts +37 -0
  167. package/src/ui/activity.ts +1 -6
  168. package/src/ui/config-edit.ts +14 -67
  169. package/src/ui/config.ts +3 -8
  170. package/src/ui/plan.ts +11 -198
  171. package/src/ui/progress.ts +3 -49
  172. package/src/ui/result.ts +18 -77
  173. package/src/ui/review-menu.ts +3 -12
  174. package/src/ui/select.ts +9 -79
  175. package/src/updater.ts +1 -1
  176. package/src/usage.ts +1 -7
  177. package/prompts/scout.md +0 -140
  178. package/src/ci/platform-estimate.ts +0 -111
  179. package/src/review/scout.ts +0 -323
@@ -14,7 +14,6 @@ export interface ReviewMenuState {
14
14
  base: string;
15
15
  pr?: number | "current";
16
16
  post: boolean;
17
- scout: boolean;
18
17
  force: boolean;
19
18
  full: boolean;
20
19
  dryRun: boolean;
@@ -48,7 +47,6 @@ export function renderReviewMenuCard(
48
47
  `Target: ${targetLabel} — Review target mode`,
49
48
  `Base: ${state.base} · Head: ${state.head} — Git diff range`,
50
49
  `Post to PR: ${state.post ? "[✓] Yes" : "[ ] No"} — Publish comments to GitHub PR`,
51
- `Scout stage: ${state.scout ? "[✓] Enabled" : "[ ] Disabled"} — Run pre-hunt reconnaissance`,
52
50
  `Force review: ${state.force ? "[✓] Yes" : "[ ] No"} — Bypass size gate thresholds`,
53
51
  `Full verification: ${state.full ? "[✓] Yes" : "[ ] No"} — Run deep refuter checks`,
54
52
  `Dry run: ${state.dryRun ? "[✓] Yes" : "[ ] No"} — Execute without posting or resolving`,
@@ -97,14 +95,11 @@ export async function runReviewMenu(
97
95
  width?: number;
98
96
  effectiveConfig?: LocalConfig;
99
97
  defaultBase?: string;
100
- defaultScout?: boolean;
101
98
  defaultPost?: boolean;
102
99
  } = {},
103
100
  ): Promise<ReviewMenuOutcome> {
104
101
  const resolvedBase =
105
102
  deps.defaultBase ?? deps.effectiveConfig?.default_base ?? "main";
106
- const resolvedScout =
107
- deps.defaultScout ?? deps.effectiveConfig?.scout ?? false;
108
103
  const resolvedPost = deps.defaultPost ?? deps.effectiveConfig?.post ?? false;
109
104
  const resolvedTarget = resolvedPost ? "pr" : "branch";
110
105
 
@@ -113,7 +108,6 @@ export async function runReviewMenu(
113
108
  head: "HEAD",
114
109
  base: resolvedBase,
115
110
  post: resolvedPost,
116
- scout: resolvedScout,
117
111
  force: false,
118
112
  full: false,
119
113
  dryRun: false,
@@ -167,7 +161,7 @@ export async function runReviewMenu(
167
161
  });
168
162
 
169
163
  let cursor = 0;
170
- const fieldCount = 7;
164
+ const fieldCount = 6;
171
165
  const actions = REVIEW_CARD_ACTIONS;
172
166
  const totalItems = fieldCount + actions.length;
173
167
  const ESC = "\x1b";
@@ -244,7 +238,6 @@ export async function runReviewMenu(
244
238
  base: state.base,
245
239
  pr: state.target === "pr" ? (state.pr ?? "current") : undefined,
246
240
  post: state.post,
247
- scout: state.scout,
248
241
  force: state.force,
249
242
  full: state.full,
250
243
  dryRun: state.dryRun,
@@ -291,12 +284,10 @@ export async function runReviewMenu(
291
284
  state.post = !state.post;
292
285
  }
293
286
  } else if (cursor === 3) {
294
- state.scout = !state.scout;
295
- } else if (cursor === 4) {
296
287
  state.force = !state.force;
297
- } else if (cursor === 5) {
288
+ } else if (cursor === 4) {
298
289
  state.full = !state.full;
299
- } else if (cursor === 6) {
290
+ } else if (cursor === 5) {
300
291
  state.dryRun = !state.dryRun;
301
292
  }
302
293
  render(false);
package/src/ui/select.ts CHANGED
@@ -1,5 +1,5 @@
1
1
  // The terminal surface's INTERACTIVE half: the keyboard-driven menu that
2
- // replaced the cost band's `[y/N]` read. ui/primitives.ts stays pure by construction;
2
+ // replaced the plain `[y/N]` read. ui/primitives.ts stays pure by construction;
3
3
  // this module is where the one unavoidable side effect — putting the user's
4
4
  // terminal into raw mode — is quarantined behind an injected reader, so every
5
5
  // decision above it (which options exist, where the cursor lands, what a key
@@ -13,7 +13,6 @@
13
13
  // keeps the old plain-text behaviour byte for byte, and the restore lives in a
14
14
  // `finally` so a throw mid-menu cannot strand a human in raw mode.
15
15
 
16
- import type { FreeRouteBilling } from "#model/free-discovery";
17
16
  import { bold, dim, log } from "./primitives";
18
17
 
19
18
  export type ConfirmResult =
@@ -82,7 +81,6 @@ interface MenuOption {
82
81
  // re-presents the menu instead of resolving it.
83
82
  action: "review" | "details" | "cancel";
84
83
  label: string;
85
- hint?: string;
86
84
  // Lowercase single characters. `y`/`n` are kept on the review/cancel rows on
87
85
  // purpose: they are what a decade of `[y/N]` prompts trained into fingers.
88
86
  shortcuts: string[];
@@ -90,11 +88,6 @@ interface MenuOption {
90
88
  }
91
89
 
92
90
  export interface ConfirmSpec {
93
- low: number;
94
- high: number;
95
- // Set only for a route the plan card already priced as free or mixed. Absent
96
- // means today's wording on both the menu hint and the plain prompt.
97
- billing?: FreeBillingMode;
98
91
  // Only true when the run was actually going to post: offering "don't post"
99
92
  // to a run that never would is an option with no meaning behind it.
100
93
  canSkipPost: boolean;
@@ -137,8 +130,7 @@ export function menuOptions(canSkipPost: boolean): MenuOption[] {
137
130
 
138
131
  export type SizeGateConfirmResult = { kind: "proceed" } | { kind: "cancel" };
139
132
 
140
- // Continue / Cancel only. No cost hint (the gate exists because we will
141
- // not guess at this bill) and no details view (the SKIP line above the
133
+ // Continue / Cancel only, and no details view (the SKIP line above the
142
134
  // menu already named the reason and the two flag hatches).
143
135
  export function sizeGateMenuOptions(): MenuOption[] {
144
136
  return [
@@ -157,55 +149,6 @@ export function sizeGateMenuOptions(): MenuOption[] {
157
149
  ];
158
150
  }
159
151
 
160
- const LABEL_WIDTH = 32;
161
-
162
- // The billing mode of an admitted route plan, derived ONCE from the same
163
- // free-step count the plan card prints (planEstimateLine in ui/plan.ts reads
164
- // this too). All steps free: nothing is billed. Some free: the band stays and
165
- // names the free share. Otherwise there is no mode, and today's wording stands.
166
- export type FreeBillingMode = "free" | "mixed";
167
-
168
- export const FREE_STEPS_NOTE = "(free steps not billed)";
169
-
170
- export function freeBillingMode(
171
- free: FreeRouteBilling | undefined,
172
- ): FreeBillingMode | undefined {
173
- if (free === undefined || free.total === 0) return undefined;
174
- if (free.free === free.total) return "free";
175
- return free.free > 0 ? "mixed" : undefined;
176
- }
177
-
178
- function band(low: number, high: number): string {
179
- return `$${low.toFixed(2)} – $${high.toFixed(2)}`;
180
- }
181
-
182
- // The menu's first-row hint. A free route shows no dollar figure at all: the
183
- // band would read as spend, which is the claim the plan card just retracted.
184
- export function costHint(
185
- low: number,
186
- high: number,
187
- billing?: FreeBillingMode,
188
- ): string {
189
- if (billing === "free") return "free route (not billed)";
190
- if (billing === "mixed") return `${band(low, high)} ${FREE_STEPS_NOTE}`;
191
- return band(low, high);
192
- }
193
-
194
- // The plain [y/N] prompt's sentence, without its question mark and answer
195
- // domain. Metered is the sentence this module always printed, byte for byte.
196
- export function spendSentence(
197
- low: number,
198
- high: number,
199
- billing?: FreeBillingMode,
200
- ): string {
201
- const range = `$${low.toFixed(2)}–$${high.toFixed(2)}`;
202
- if (billing === "free") {
203
- return `Run this review on a free route (token-derived ${range}, not billed)`;
204
- }
205
- const note = billing === "mixed" ? ` ${FREE_STEPS_NOTE}` : "";
206
- return `Spend an estimated ${range} on this review${note}`;
207
- }
208
-
209
152
  // The frame, as lines. Built unstyled and painted per line, same rule as the
210
153
  // plan card: the pointer is plain text either way, so a piped run reads
211
154
  // correctly with no escapes at all.
@@ -217,10 +160,7 @@ export function renderMenu(
217
160
  return options.map((option, i) => {
218
161
  const active = i === cursor;
219
162
  const head = active ? "❯ " : " ";
220
- const body = option.hint
221
- ? `${option.label.padEnd(LABEL_WIDTH)}${option.hint}`
222
- : option.label;
223
- const line = ` ${head}${body}`;
163
+ const line = ` ${head}${option.label}`;
224
164
  if (!styles) return line;
225
165
  return active ? bold(line, true) : dim(line, true);
226
166
  });
@@ -265,11 +205,6 @@ export async function runConfirm(
265
205
  if (!reader.raw) return runPlainConfirm(spec, reader, io);
266
206
 
267
207
  const options = menuOptions(spec.canSkipPost);
268
- const withHint = options.map((option, i) =>
269
- i === 0
270
- ? { ...option, hint: costHint(spec.low, spec.high, spec.billing) }
271
- : option,
272
- );
273
208
  let cursor = 0;
274
209
  // Repainting in place needs cursor-movement escapes, which are exactly what
275
210
  // must never reach a pipe. When styling is off the frames are simply
@@ -279,7 +214,7 @@ export async function runConfirm(
279
214
 
280
215
  const draw = (): void => {
281
216
  if (repaint && drawn > 0) io.write(`${ESC}[${drawn}A`);
282
- const lines = renderMenu(withHint, cursor, spec.styles);
217
+ const lines = renderMenu(options, cursor, spec.styles);
283
218
  for (const line of lines) {
284
219
  if (repaint) io.write(`${ESC}[2K`);
285
220
  io.line(line);
@@ -313,19 +248,15 @@ export async function runConfirm(
313
248
  return { kind: "cancel" };
314
249
  }
315
250
  if (key.type === "up" || key.type === "down") {
316
- cursor = moveCursor(
317
- cursor,
318
- key.type === "up" ? -1 : 1,
319
- withHint.length,
320
- );
251
+ cursor = moveCursor(cursor, key.type === "up" ? -1 : 1, options.length);
321
252
  draw();
322
253
  continue;
323
254
  }
324
255
  const chosen =
325
256
  key.type === "enter"
326
- ? withHint[cursor]
257
+ ? options[cursor]
327
258
  : key.type === "char"
328
- ? withHint.find((option) => option.shortcuts.includes(key.char))
259
+ ? options.find((option) => option.shortcuts.includes(key.char))
329
260
  : undefined;
330
261
  if (chosen === undefined) continue;
331
262
  if (chosen.action === "details") {
@@ -357,7 +288,7 @@ async function runPlainConfirm(
357
288
  io: ConfirmIo,
358
289
  ): Promise<ConfirmResult> {
359
290
  io.line();
360
- io.write(`${spendSentence(spec.low, spec.high, spec.billing)}? [y/N] `);
291
+ io.write("Run this review? [y/N] ");
361
292
  const answer = (await reader.read())?.trim().toLowerCase() ?? "";
362
293
  io.line();
363
294
  return answer === "y" || answer === "yes"
@@ -437,8 +368,7 @@ export async function confirmReview(
437
368
  }
438
369
 
439
370
  // Sibling of runConfirm for the size-gate override. Same reader/io contract,
440
- // same unanswered-is-no rule, no details view and no dollar hint — guessing
441
- // at this bill is the thing the gate refuses to do.
371
+ // same unanswered-is-no rule, no details view.
442
372
  export async function runSizeGateConfirm(
443
373
  styles: boolean,
444
374
  reader: KeyReader,
package/src/updater.ts CHANGED
@@ -55,7 +55,7 @@ export function detectInstallMethod(
55
55
  ): InstallMethod {
56
56
  const home = options.home ?? os.homedir();
57
57
  const execPath = options.execPath ?? process.execPath;
58
- const version = options.version ?? "0.6.1";
58
+ const version = options.version ?? "0.8.0";
59
59
  const exists = options.exists ?? existsSync;
60
60
 
61
61
  const isSource =
package/src/usage.ts CHANGED
@@ -6,11 +6,9 @@ export interface SessionUsage {
6
6
  tokens_in: number;
7
7
  tokens_out: number;
8
8
  tokens_total: number;
9
- cost_usd_est: number;
10
9
  }
11
10
 
12
11
  interface RawSessionResult {
13
- total_cost_usd?: number;
14
12
  usage?: {
15
13
  input_tokens?: number;
16
14
  output_tokens?: number;
@@ -19,7 +17,7 @@ interface RawSessionResult {
19
17
  };
20
18
  }
21
19
 
22
- // Session-total tokens/cost is the MUST (Rec-R3); per-agent attribution is a
20
+ // Session-total tokens is the MUST (Rec-R3); per-agent attribution is a
23
21
  // documented SHOULD/best-effort gap — `--output-format json` only exposes the
24
22
  // session aggregate (verified against a live headless run), never a
25
23
  // per-hunter/refuter breakdown, so this slice does not populate it.
@@ -35,7 +33,6 @@ export function parseUsage(rawStdout: string, wall_ms: number): SessionUsage {
35
33
  tokens_in: 0,
36
34
  tokens_out: 0,
37
35
  tokens_total: 0,
38
- cost_usd_est: 0,
39
36
  };
40
37
  }
41
38
  const usage = parsed.usage ?? {};
@@ -49,7 +46,6 @@ export function parseUsage(rawStdout: string, wall_ms: number): SessionUsage {
49
46
  tokens_in,
50
47
  tokens_out,
51
48
  tokens_total: tokens_in + tokens_out,
52
- cost_usd_est: parsed.total_cost_usd ?? 0,
53
49
  };
54
50
  }
55
51
 
@@ -59,7 +55,6 @@ export function zeroUsage(): SessionUsage {
59
55
  tokens_in: 0,
60
56
  tokens_out: 0,
61
57
  tokens_total: 0,
62
- cost_usd_est: 0,
63
58
  };
64
59
  }
65
60
 
@@ -72,6 +67,5 @@ export function sumUsage(a: SessionUsage, b: SessionUsage): SessionUsage {
72
67
  tokens_in: a.tokens_in + b.tokens_in,
73
68
  tokens_out: a.tokens_out + b.tokens_out,
74
69
  tokens_total: a.tokens_total + b.tokens_total,
75
- cost_usd_est: a.cost_usd_est + b.cost_usd_est,
76
70
  };
77
71
  }
package/prompts/scout.md DELETED
@@ -1,140 +0,0 @@
1
- ---
2
- name: pr-hero-scout
3
- description: Diff-only triage pass that emits unverified leads biasing the hunters' scan.
4
- ---
5
-
6
- You are the engine-owned scout. You read ONE diff and nothing else. You have no
7
- repository access and no tools: you cannot open a file, grep, follow a symbol,
8
- or confirm anything you suspect. Everything you emit is an unverified
9
- suspicion, and you must never write as though you had checked it.
10
-
11
- Your job is to point the reviewers at the places in this diff most likely to
12
- hide a defect. You do not decide whether a defect is actually there — reviewers
13
- with repository access do that, and they read the whole diff regardless of what
14
- you say. A lead you get wrong costs a reviewer a moment of attention on code
15
- they were going to read anyway. A place you fail to point at is not excluded
16
- from review.
17
-
18
- That asymmetry sets your bias: **within your cap, prefer recall.** A suspicion
19
- you can state in one sentence is worth emitting even when you cannot tell
20
- whether it holds. Do not withhold a lead because you cannot prove it — proving
21
- it is not your job and you are not equipped for it.
22
-
23
- The cap is the other half. Leading on every hunk is the same as leading on
24
- nothing. Emit **at most 12 leads**, at most **3 per file**, and spend them on
25
- the places where the change is least self-evidently correct. When more places
26
- deserve a lead than you have budget for, break the tie by consequence — what
27
- goes wrong if you are right, not how unsure you feel. Both limits are applied
28
- after you answer, by keeping your leads **in the order you wrote them**: a
29
- fourth lead on one file is discarded, never weighed against the three before
30
- it. So within a file, write the worst consequence first. A diff whose changes
31
- are all self-evident is a real and expected outcome: return `{"leads":[]}`
32
- rather than manufacturing suspicion.
33
-
34
- ## What counts as a lead
35
-
36
- Any location in the changed code whose correctness is not settled. Ask three
37
- questions of every change, in this order. They are ordered by how much they
38
- yield, and the later ones must not crowd out the first.
39
-
40
- **1. Is this wrong on its own terms?** Read the added lines as they stand and
41
- work out what they do for every input they can receive — not the input the
42
- author had in mind. Most defects are fully decidable from the lines in front of
43
- you, and they are the ones a reader skims past because the code looks
44
- reasonable. This is where the majority of your leads should come from.
45
-
46
- **2. What does this change assume about the REST OF THIS DIFF?** A patch is one
47
- change, not a list of files. An enum gains a case in one hunk and a switch over
48
- it sits unmodified in another; a producer changes shape here and its second
49
- consumer is three files down, untouched. Those relations are decidable from
50
- what you were handed, and they are leads.
51
-
52
- **3. What does it assume about code you cannot see?** Callers, siblings and
53
- concurrent execution outside the patch. Legitimate, and the smallest of the
54
- three. A lead list made entirely of "this depends on something not in the diff"
55
- is a scout that read the patch as a list of external dependencies instead of as
56
- code.
57
-
58
- ### Three places a defect hides best
59
-
60
- These are the classes a reviewer's eye slides over, and they are most of the
61
- reason you exist. Check every diff for all three.
62
-
63
- - **Deletions.** Additions look dangerous; deletions look like cleanup. But a
64
- removed field, config flag, default, branch or interface member can change
65
- behaviour while the code still compiles and the tests still pass. Of every
66
- deletion ask: who depended on this, and what used to be true that is not
67
- anymore?
68
- - **One side of a boundary updated.** The patch changes one producer of two,
69
- one handler of an enum, one implementation of an interface, one caller of a
70
- changed signature — and leaves its sibling as it was. Neither side breaks on
71
- its own, which is exactly why nothing catches it. When both sides are in this
72
- diff you can see it outright; when only one is, name the sibling you would
73
- expect to have moved.
74
- - **Silent behaviour changes.** The signature holds and the name holds, but the
75
- behaviour does not: an error now swallowed that used to propagate, a miss now
76
- treated as a failure or a failure as a miss, a default that changed value.
77
-
78
- ### And the ordinary ways code is wrong
79
-
80
- - A comparison or a piece of arithmetic bounds one direction and leaves the
81
- other open, so the failing input is the one that still passes the check — a
82
- quantity arriving negative, reversed, out of range, or larger than the code
83
- assumes. **Run this one actively instead of waiting to notice it:** take
84
- every comparison and every subtraction in the added lines and name the input
85
- that makes it come out the way the author wants when it should not. If you
86
- cannot name one, move on. If you can, you have a lead.
87
- - A value, unit, bound, key, order or lifetime is assumed rather than shown.
88
- - An edge the new code does not visibly handle: empty, absent, zero, negative,
89
- duplicate, overlong, already-done, failed, or arriving twice.
90
- - The stated intent of the change and what the code now does appear to diverge.
91
-
92
- ### What is NOT a lead
93
-
94
- This half is where a scout earns the right to be read at all.
95
-
96
- - **A problem in code this patch does not change.** Context lines are there to
97
- be read, not audited. If the change did not create it, touch it, or newly
98
- expose it, it is not yours.
99
- - Style, naming, formatting, documentation quality, or a preference about how
100
- it might have been written. You are not a linter and you are not a critic.
101
- - **What is not tested.** A missing case, a thin assertion, an untested branch
102
- — none of those is yours, and neither is a test that could be written better.
103
- A test that asserts something FALSE about the code it exercises is a
104
- different animal, and that one is fair game.
105
- - A restatement of what the diff plainly does, "this file changed", or "this
106
- function is complex".
107
- - **The same concern twice.** One suspicion earns ONE lead even when it repeats
108
- in three files: cite the site where the consequence is worst and name the
109
- others inside that same `why`. A second lead reading "the same issue as the
110
- file above" spends a slot on a sentence instead of on a suspicion, and your
111
- slots are the only thing that makes you selective at all.
112
- - The same location twice.
113
-
114
- ## The diff is not necessarily source code
115
-
116
- A patch may change a shell script, a runbook or design note in Markdown, a CI
117
- workflow, a Dockerfile, a database migration, a lockfile, an infrastructure
118
- template, or a configuration file. **Defects in those are as real as defects in
119
- application code and you scan them with the same eye** — a command whose
120
- failure is silent, a loop with no bound, a permission widened, a variable that
121
- may be unset, a guard that no longer runs. Prose inside a Markdown file is not
122
- a lead; an executable instruction inside it is.
123
-
124
- ## Coordinates
125
-
126
- - `path` — exactly as the diff names the file, without the `a/` or `b/` prefix.
127
- - `line` — the line number **in the new version of the file**. Count it from
128
- the hunk header `@@ -old,n +new,m @@`: the first line after that header is
129
- line `new`, and every context line and every added (`+`) line advances the
130
- count by one. Removed (`-`) lines do not. For a change whose suspicious part
131
- spans several lines, cite the first of them. Getting this number wrong makes
132
- a correct suspicion useless, so derive it, do not estimate it.
133
- - `why` — ONE sentence, at most 240 characters. Name the concrete behaviour at
134
- risk and where the reviewer should start. Not a verdict, not a severity, not
135
- a fix, and never a hedge — "consider checking X" is noise, and a `why` whose
136
- whole content is "this depends on something outside the diff" is an
137
- observation rather than a suspicion. The register to aim at: *this deletion
138
- looks suspicious*; *this enum case isn't handled in the sibling file*; *this
139
- error path is silently swallowing failures*. If it needs a paragraph, it is a
140
- finding in disguise and you are not the one making it.
@@ -1,111 +0,0 @@
1
- // The review cost the Crysk platform estimated for this run (cryskai/platform#12).
2
- // The connect step asks for it with `X-Crysk-Capabilities: estimate` and hands it
3
- // to the run step as `CRYSK_PLATFORM_ESTIMATE`, where the CI budget gate reads it
4
- // instead of the engine's own band (cryskai/crysk#138): the platform knows the
5
- // setup's models, their catalog prices and what its recent reviews cost; the
6
- // band only knows the diff's size and prices every model as Claude.
7
- //
8
- // The platform already checked it, and the engine checks again: a value that is
9
- // not a usable range is dropped, and the band decides, as without a platform.
10
-
11
- export type PlatformEstimateBasis = "history" | "catalog";
12
-
13
- export interface PlatformEstimate {
14
- readonly usdLow: number;
15
- readonly usdHigh: number;
16
- // `history`: the setup's recent reviews scaled to this pull request.
17
- // `catalog`: the catalog prices of the setup's models.
18
- readonly basis: PlatformEstimateBasis;
19
- }
20
-
21
- export const PLATFORM_ESTIMATE_CAPABILITY = "estimate";
22
-
23
- const usable = (value: unknown): value is number =>
24
- typeof value === "number" && Number.isFinite(value) && value >= 0;
25
-
26
- // The `/runs/token` field, `{ usd_low, usd_high, basis }`, or undefined when it
27
- // is absent or not a usable range.
28
- export function parsePlatformEstimate(
29
- value: unknown,
30
- ): PlatformEstimate | undefined {
31
- if (typeof value !== "object" || value === null || Array.isArray(value)) {
32
- return undefined;
33
- }
34
- const {
35
- usd_low: low,
36
- usd_high: high,
37
- basis,
38
- } = value as Record<string, unknown>;
39
- if (!usable(low) || !usable(high) || low > high) return undefined;
40
- if (basis !== "history" && basis !== "catalog") return undefined;
41
- return { usdLow: low, usdHigh: high, basis };
42
- }
43
-
44
- // What the connect step writes to its `estimate` output: one line of JSON in the
45
- // response's own field names, numbers and an enum only.
46
- export function serializePlatformEstimate(estimate: PlatformEstimate): string {
47
- return JSON.stringify({
48
- usd_low: estimate.usdLow,
49
- usd_high: estimate.usdHigh,
50
- basis: estimate.basis,
51
- });
52
- }
53
-
54
- // The run step's read of `CRYSK_PLATFORM_ESTIMATE`. Empty (a standalone run, an
55
- // older platform, nothing priceable) or malformed reads as no estimate.
56
- export function readPlatformEstimate(
57
- env: Readonly<Record<string, string | undefined>>,
58
- ): PlatformEstimate | undefined {
59
- const text = env.CRYSK_PLATFORM_ESTIMATE;
60
- if (text === undefined || text === "") return undefined;
61
- try {
62
- return parsePlatformEstimate(JSON.parse(text));
63
- } catch {
64
- return undefined;
65
- }
66
- }
67
-
68
- // The estimate as the plan and the CI budget gate read a cost band (report.ts's
69
- // CostEstimate, structurally: this module imports nothing).
70
- export function platformCostEstimate(estimate: PlatformEstimate): {
71
- low: number;
72
- high: number;
73
- basis: string;
74
- platform: PlatformEstimateBasis;
75
- } {
76
- return {
77
- low: estimate.usdLow,
78
- high: estimate.usdHigh,
79
- basis:
80
- estimate.basis === "history"
81
- ? "The Crysk platform's estimate, from this setup's recent reviews scaled to this pull request."
82
- : "The Crysk platform's estimate, from the catalog prices of this setup's models.",
83
- platform: estimate.basis,
84
- };
85
- }
86
-
87
- // The estimate the run gates and plans on. The platform prices the whole pull
88
- // request from the setup's own models, before the engine knows how much of it
89
- // this run reviews. A narrowed run (a re-review whose discovery reads only what
90
- // changed since the last review, or one that only verifies earlier findings)
91
- // passes `fullBand`, the band for a full review of the pull request: the
92
- // platform's figure is scaled by the share of that work the run does
93
- // (band / fullBand, at most 1). The band measures the share only; its prices,
94
- // which treat every model as Claude, never replace the platform's. A full review
95
- // uses the platform's figure; without one the band decides, as before.
96
- export function runCostEstimate<Band extends { high: number }>(
97
- platform: PlatformEstimate | undefined,
98
- band: Band,
99
- fullBand: { high: number } | undefined,
100
- ): Band | ReturnType<typeof platformCostEstimate> {
101
- if (platform === undefined) return band;
102
- const estimate = platformCostEstimate(platform);
103
- if (fullBand === undefined || !(fullBand.high > 0)) return estimate;
104
- const share = Math.min(1, Math.max(0, band.high / fullBand.high));
105
- return {
106
- ...estimate,
107
- low: estimate.low * share,
108
- high: estimate.high * share,
109
- basis: `${estimate.basis} Scaled to the ${Math.round(share * 100)}% of a full review this run does.`,
110
- };
111
- }