213 shown · click any identifier for its record and every place it appears

1. AC-001Architecture must separate the experience layer from the agent runtime10
2. AC-002Every PDLC claim on the site must be verifiable; no simulated delivery metrics10
3. AC-C10It works without motion6
4. AC-C11It is fast on a cold visit1
5. AC-C12Nothing on the delivery surface is decorative3
6. ADR-0001Split the experience layer from the agent runtime4
7. ADR-0002The product name is build-time configuration, not source9
8. ADR-0003Host the agent runtime on Hugging Face Spaces (free CPU)15
9. ADR-0004No local neural embedding model in the container13
10. ADR-0005Vectors from different embedding models are never compared23
11. ADR-0006Human-approval nodes are terminal by construction19
12. ADR-0007Rate limiting and the spend guard fail closed23
13. ADR-0008Similarity thresholds belong to the embedder, not to the cache11
14. ADR-0009Models are trained offline and served without their training framework30
15. ADR-0010Training labels are true by construction, never model-generated11
16. ADR-0011No approximate-nearest-neighbour index until the corpus reaches ~5,000 chunks6
17. ADR-0012The agent runtime runs on Northflank, not Hugging Face Spaces12
18. ADR-0013The trained classifier stays out of the live evidence gate13
19. ADR-0014One Vercel project, connected to git4
20. BR-001Triage an incident, produce a hypothesis with cited evidence3
21. BR-002Demonstrate PDLC/SDLC/CI-CD/Agile/Scrum verifiably4
22. BR-003Continue operating when a provider fails or rate-limits6
23. BR-004Refuse when evidence does not support an answer3
24. BR-005Record a full, inspectable execution trace per run7
25. BR-006Block high-risk actions pending human approval2
26. BR-007Attribute tokens and cost to each run2
27. BR-008Retain session and cross-incident memory2
28. BR-009Evaluate quality against a fixed golden set2
29. BR-010Reduce redundant model calls via semantic caching3
30. BR-011Present own architecture, decisions and delivery record2
31. D-001Refusal gate marked 150/265 unanswerable questions answerable — a 56.6% false-answer rate reported at the Sprint 2 gate as zero16
32. D-002Evidence gate answered a value-demanding question the corpus never answers, scoring 8.852
33. D-003Re-ranker experiment returned a null result that was two bugs: a NaN checkpoint and a saturated metric4
34. D-004Seed loader destroyed one embedding model's vectors when reloading under another2
35. D-005Retry loop re-sent an identical prompt, so it could never succeed8
36. D-006Semantic cache served the previous answer to a correction retry (0.886 similarity vs 0.60 threshold)13
37. D-007RUN_BUDGET_USD=0 killed every run mid-stream with an unhandled error8
38. D-008CSP blocked React hydration in development; every button was inert HTML7
39. D-009The console displayed assessments the governance layer had refused to emit8
40. D-010Spend reservation priced against providers[0], under-reserving 4x when failover reached a costlier provider11
41. D-011A comment inside a backslash continuation truncated the Semgrep invocation; the scan ran without its exclusion, printed success, and the shell exited 127 on the orphaned flag8
42. D-012No body-size limit on the run endpoint; a 200KB body reached the agent and returned 502 instead of being refused3
43. D-013The rate-limit pen test could not fail: it sent 8 requests against a limit of 20, and its pass condition accepted all(c >= 400), so a service that was DOWN reported as correctly rate limited19
44. D-014Traceability statuses drifted in both directions: 4 rows used statuses the legend never defined (Done (design)/(gate)/(decision)) which the delivery page counted as done, and 1 row sat at Planned while its test had passed for four sprints12
45. D-015The first README checker could not fail: it searched for each figure as a substring of the whole file, so changing `Commits \12
46. D-016Sprints 6 and 7 were worked and shipped with no planning ceremony and no plan document; the sprint numbers existed only in defect-log entries, and a Sprint 5 retrospective commitment to raise exactly this was never honoured12
47. D-017ADR-0003 placed the agent runtime on Hugging Face Docker Spaces "because it is free"; Docker Spaces are PRO-only, so three sprints of deployment work targeted a platform that cannot host it at $013
48. D-018CI never built the web application. A Dependabot PR taking Next 15→16 reported 10/10 green while the production build failed on all seven pages; no job had ever run next build7
49. D-019Two governance documents (PROJECT_RECORD.html, SPRINT_08_PLAN.md) asserted the web app was undeployed and Sprint 8 was blocked on credentials for a full week after the real deploy landed; caught only because the Product Owner quoted the stale text back and asked "true?"15
50. D-02020 of 45 Planned rows in the traceability matrix (BR-001/004/006/007/008/009/010, FR-002/005/006/007/010/014/015/016/020/025/026/027/029) had real, passing, CI-green tests months before the row was updated — a recurrence of D-014's exact root cause at 20x the scale, caught only by an explicit read-the-code-not-the-doc audit requested by the Product Owner ("I want all of the 45 to be done"). PROJECT_RECORD.html (the same document D-019 was found in) independently carried the same stale 13 done / 45 Planned figures in three places, hand-typed rather than generated13
51. D-021Every scene on the landing page rendered at opacity: 0 without JavaScript. Scene revealed itself with an IntersectionObserver and a visible state starting false, so the server-rendered markup carried an inline opacity: 0 — measured with JS disabled, all 7 scenes computed to 0. The component's own docstring claimed "content is present in the DOM and legible with animation removed"; it was present and invisible. Reached production and sat there for four sprints7
52. D-022The public architecture diagram (SystemMap.tsx) labelled the runtime "hugging face space" for a week after ADR-0012 moved it to Northflank — the same stale-claim class as D-017 and D-019, on the surface a reviewer is most likely to read4
53. D-023All 12 "read the decision" links on the public architecture surface pointed at the ADR directory rather than at a record: the href was built by concatenating a path ending /adr/ with a filename the derived record never carried. Twelve dead links, each rendering as an ordinary underlined title. Found by the first e2e assertion that every decision link resolves to a .md4
54. D-024The memory panel rendered the session id, which is read from sessionStorage during render — "" on the server, a real id in the browser. Every page carrying the console threw React #418 (hydration mismatch) in production. Reached main; caught by CI, on a build CI made itself5
55. D-025The local e2e suite reported green against a STALE build twice: Playwright's reuseExistingServer: !process.env.CI reused a server left running from an earlier build, so the run exercised the previous binary. It masked a fix that had not been rebuilt, then masked D-024 entirely — CI caught that one only because it builds from scratch3
56. D-026Two labels on the public architecture diagram were struck through by the dashed lines they sat on. The masking plates were sized in SVG user units, so a width that covered the text on macOS was too short on Linux and at mobile scale — the fix for the first version broke on a platform the first version happened to pass on2
57. D-027The public architecture diagram asserted three things about the system that were not true. The arrow pointing at Upstash had no source node — its tail began in empty space, because the only caller is the edge rate limiter and the node was drawn in the runtime's data column. The node was labelled redis · vector; Upstash Vector appears exactly once in the repository, in training/benchmark_vector_store.py, as the managed-store comparison arm for ADR-0011, and has never served a request. And an edge left the Provider chain for Neon, which never opens a database connection — nothing in router.py imports the db module. Reported by the Product Owner reading the deployed page4
58. D-028The role tabs on the story surface gave no hover feedback at all once selected: .role-tab:hover and .role-tab[data-on="true"] have identical specificity (0,2,0), so the later rule took back both color and background and the pointer sat over a control that showed no sign of being one. Selection was also carried by colour alone — color, background and border-color, every one of which a forced-colors mode overrides — so in Windows High Contrast the chosen role was indistinguishable from the other ten, and the page could not tell you whose story you were reading. The global :focus-visible additionally set border-radius: 4px, squaring off the 999px pill for keyboard users only. Reported by the Product Owner1
59. D-029The label of a role tab was unreadable for ~60ms every time it was selected. transition: color, background interpolates the two independently, and the path from unselected (#a1a1a8 on #0b0b0d) to selected (#000 on #f5f5f7) runs the text toward black while the background is still near-black. Measured through the fade: contrast 7.66 → 5.39 → 1.71 → 8.62 → 19.29. Reported by the Product Owner as "the button does not display the text that is selected", with a screenshot that had caught the surface mid-fade3
60. D-030The traceability matrix cites 26 sprint-story identifiers that no sprint plan ever defines — S1-DATA, S1-FEED, S1-SEED, S2-CACHE, S2-CHAOS, S2-EVAL, S2-GOV, S2-GUARD, S2-RAG, S2-ROUTER, S2-SPEND, S3-CITE, S3-MEM, S3-RCA, S3-STREAM, S3-TRACE, S3-TRIAGE, S4-PERF, S4-UX, S5-ARCH, S5-DELIV, S6-DELIV, S6-REVIEW, S9-DATA, S9-GOV, S9-KT. Sprints 5 onward list their stories in a table; sprints 1-4 never did, so the matrix's story column points at records that do not exist. Three code comments also cite FR-036, FR-040 and FR-046, none of which appear in the matrix — they are the far ends of range shorthand (FR-030..FR-036) written without checking the range existed. Found within minutes of building the lexicon (FR-034), which is the argument for it1
61. FR-001Product name changeable without code changes8
62. FR-002Simulated production estate, deterministically seeded4
63. FR-003Incident feed on schedule and on demand11
64. FR-004Live triage run streamed to the client5
65. FR-005Cited evidence panel; unsupported claims marked4
66. FR-006Explicit refusal on insufficient evidence5
67. FR-007Human approval gate; approval is terminal6
68. FR-008Session memory, visible to the user7
69. FR-009Postmortem drafting from a completed run9
70. FR-010Deterministic router with time-boxed provider disabling3
71. FR-011Visitor-triggered provider failure injection20
72. FR-012Semantic cache with visible hit rate and spend avoided5
73. FR-013Execution trace viewer3
74. FR-014Cost attribution and pre-flight spend guard3
75. FR-015Evaluation harness over the golden set5
76. FR-016Known-limitation probe suite that warns every run5
77. FR-017Cinematic scroll-driven product narrative2
78. FR-018Reduced-motion and mobile comprehension paths2
79. FR-019Interactive architecture view4
80. FR-020Live CI/CD status from the GitHub API8
81. FR-021Requirements traceability rendered publicly5
82. FR-022Sprint and velocity record from real commit history4
83. FR-023Decision records rendered with context and consequences7
84. FR-024Postmortems for real defects hit during the build4
85. FR-025Change risk review workload on the same orchestration graph4
86. FR-026Labelled evaluation dataset, labels true by construction3
87. FR-027Statistical evaluation of the refusal gate4
88. FR-028Trained, calibrated evidence-sufficiency classifier8
89. FR-029Cross-encoder re-ranker trained in PyTorch, served via ONNX5
90. FR-030Approximate nearest neighbour benchmark12
91. FR-031The synthetic dataset published in full5
92. FR-032The governance record published4
93. FR-033Handover document serving non-technical and architect readers alike7
94. FR-034Every identifier the project cites is followable to its own record and every place it appears3
95. FR-036cited 2 times, but no record in the repository defines it2
96. FR-040cited 2 times, but no record in the repository defines it2
97. FR-046cited 2 times, but no record in the repository defines it2
98. IMP-01Docker unavailable on host — container cannot be validated locally10
99. IMP-02Product name undecided — blocks branding assets, not code2
100. IMP-03npm workspaces glob pointed at empty directories1
101. IMP-04No managed Postgres or Redis credentials. Host has no local Postgres and no Docker (IMP-01), so nothing can be executed against a real database locally.16
102. IMP-05Upstash Vector index not provisioned1
103. IMP-06Data stores are in three regions, none of them the one the agent will run in2
104. IMP-07the video budget and the visual are in direct tension1
105. IMP-08the evidence is unflattering1
106. IMP-09RESOLVED1
107. IMP-10the container image has never been built1
108. IMP-11six font families is a real payload2
109. NFR-001Effort directed at delivery, not deliberation4
110. NFR-002Zero infrastructure cost18
111. NFR-003First meaningful paint under 2.5s on cold 4G7
112. NFR-004Public endpoint survives untrusted traffic without unbounded cost9
113. NFR-005Runtime holds no persistent local state16
114. P-1Call the runtime directly, bypassing the BFF5
115. P-2Replay a captured bearer token from the browser7
116. P-3Exceed the per-IP limit, then exceed it from a rotated address13
117. P-4Take Redis offline mid-scan1
118. P-5Prompt injection through the incident body: instruct the agent to ignore its evidence6
119. P-6Prompt injection via a citation marker pointing outside the evidence set1
120. P-7Submit a 4,000-character body of repeated tokens12
121. P-8Hold 50 concurrent SSE streams open without reading1
122. P-9Approve a run created by a different session1
123. P-10Request a trace containing a secret-shaped attribute2
124. P-11Force every provider to fail1
125. P-12Drive spend past the per-run ceiling1
126. P-13security headers present on a live response — pass, covering T-124
127. S1-DATAcited 2 times, but no record in the repository defines it2
128. S1-FEEDcited 2 times, but no record in the repository defines it2
129. S1-SEEDcited 3 times, but no record in the repository defines it3
130. S2-CACHEcited 3 times, but no record in the repository defines it3
131. S2-CHAOScited 2 times, but no record in the repository defines it2
132. S2-EVALcited 4 times, but no record in the repository defines it4
133. S2-GOVcited 3 times, but no record in the repository defines it3
134. S2-GUARDcited 2 times, but no record in the repository defines it2
135. S2-RAGcited 3 times, but no record in the repository defines it3
136. S2-ROUTERcited 3 times, but no record in the repository defines it3
137. S2-SPENDcited 3 times, but no record in the repository defines it3
138. S3-ANNpgvector HNSW vs IVFFlat vs exact vs managed vector store: recall@k, latency percentiles, build time, memory4
139. S3-CITEcited 2 times, but no record in the repository defines it2
140. S3-DATAQuestion generator: ~600 examples, labels true by construction, document-level splits3
141. S3-MEMcited 3 times, but no record in the repository defines it3
142. S3-MODELFeature extraction, logistic-regression baseline, gradient-boosted challenger, probability calibration, model card3
143. S3-RCAcited 2 times, but no record in the repository defines it2
144. S3-RERANKCross-encoder fine-tuned in PyTorch, ONNX export, parity test, latency budget3
145. S3-STATSROC and PR curves, Youden's J operating point, bootstrap CIs, McNemar between configurations, power analysis3
146. S3-STREAMcited 2 times, but no record in the repository defines it2
147. S3-TRACEcited 3 times, but no record in the repository defines it3
148. S3-TRIAGEcited 2 times, but no record in the repository defines it2
149. S4-ADJUDICATEGrounding adjudicator for the ambiguous band, one cheap model call2
150. S4-CITECitation extraction and the uncited-claim verify loop2
151. S4-EVALEval harness: golden/core blocking, golden/probe warning2
152. S4-GRAPHLangGraph orchestrator; await_approval terminal by construction2
153. S4-PERFcited 2 times, but no record in the repository defines it2
154. S4-SPENDSpend guard at the single chokepoint, priced pre-flight, ledgered after2
155. S4-UXcited 4 times, but no record in the repository defines it4
156. S4-WORKLOADWorkload abstraction: incident triage and change review on ONE graph3
157. S5-APIFastAPI surface on the agent runtime; bearer auth, constant-time compare2
158. S5-APPROVEApproval gate as UI; decision creates the continuation run2
159. S5-ARCHcited 3 times, but no record in the repository defines it3
160. S5-BFFNext.js 16 on Vercel; SSE proxy; per-IP rate limit via Redis, failing closed2
161. S5-DELIVcited 7 times, but no record in the repository defines it7
162. S5-MEMSession memory panel: what was written, what was recalled2
163. S5-SMOKEAssembled-system test1
164. S5-STREAMLive triage: node events, retrieval hits, citations, refusals as they happen2
165. S5-TRACEOpenTelemetry spans across both runtimes, rendered as a waterfall2
166. S6-A11YReduced motion and small screens lose nothing1
167. S6-COPYLanding copy states the product, not the author1
168. S6-DELIVcited 2 times, but no record in the repository defines it2
169. S6-HEROScroll-scrubbed hero from stock footage1
170. S6-MEDIAReproducible media pipeline1
171. S6-MOTIONMotion system in design tokens, not ad-hoc transitions1
172. S6-NAVEvery surface reachable from every surface1
173. S6-REVIEWcited 2 times, but no record in the repository defines it2
174. S7-ARCH/architecture — request path and all ADRs1
175. S7-DERIVEderive-surfaces.mjs — one source for both pages1
176. S7-GOVA sprint number cannot precede its plan1
177. S7-MOBILEBoth surfaces usable on a phone1
178. S7-REL/reliability — measured error rates against their budgets1
179. S7-WEAKPublish the checks that are currently FAILING1
180. S8-E2EPen tests green against the deployed system5
181. S8-GUARDGuard-of-the-guard tests (Sprint 7 improvement 1)2
182. S8-HOSTAgent runtime live on Northflank3
183. S8-LOADLoad behaviour within the free tier3
184. S8-OBSRunbook and observability3
185. S8-THREATThreat model review against the deployed surface3
186. S8-VERCELExperience layer live on Vercel3
187. S9-CONTEXTWhy it was built, what the scope is, what is missing2
188. S9-DATAcited 2 times, but no record in the repository defines it2
189. S9-FINDEvery identifier the project cites is followable2
190. S9-GOVcited 2 times, but no record in the repository defines it2
191. S9-KTcited 4 times, but no record in the repository defines it4
192. S9-MARKA favicon that works at 16px1
193. S9-MOTIONScroll-driven motion system1
194. S9-ROLERole-perspective chooser inside /story2
195. S9-STORY/story — the project explained to anyone3
196. S9-TYPEThree typographic voices, one per surface family4
197. T-1Abuser drives unbounded model spend16
198. T-2Provider keys exfiltrated4
199. T-3Agent runtime called directly, bypassing rate limits9
200. T-4Prompt injection via retrieved corpus4
201. T-5Prompt injection via incident description4
202. T-6Chaos endpoint used to degrade the demo for others9
203. T-7Trace or approval records forged4
204. T-8Delivery record misrepresented2
205. T-9Session correlation from IP8
206. T-10Dependency supply chain2
207. T-11Model output triggers unintended action7
208. T-12The inter-service token reaches the browser14
209. T-13The runtime is called directly, bypassing the BFF's rate limit2
210. T-14SSE connections held open to exhaust the runtime6
211. T-15A run's free-text body used to smuggle instructions15
212. T-16A secret reaches a trace attribute and is rendered5
213. T-17Approval forged or replayed6