# noCV engineering task library Content version 5 Fictional engineering practice briefs. Starter repositories, fixtures, automated grading, and verified ownership are not included. Tests, patches, and runbooks are requested deliverables. They become Outcome Evidence only through a qualified Mission and immutable Evidence IDs. Independent adaptation must be observed under a declared verification policy and cite immutable Evidence IDs. Completing a planning ticket establishes no Ownership Evidence. ## BTRIAGE — Incident assistant with read-only tools A fictional on-call team wants an assistant that joins alerts, deployment notes, and runbooks. Incomplete telemetry and hostile log content make unsupported conclusions dangerous. **Field:** Applied AI. **Suggested stack:** TypeScript, AiEvaluationProvider, JSON Schema. **Engineer value:** Practice constrained tool use, operational uncertainty, and traceable AI outputs. **Company value:** Produce concise incident drafts that preserve source traceability and operator control. **Delivery agreement:** Deliver a local read-only prototype; it must execute no remediation and contact no external service. ### Setup prerequisites - Author synthetic alerts, deployment records, and runbooks plus deterministic tool and model doubles. ### Scope incident inputs Authorize and normalize incident records. #### BTRIAGE-101 — Define a common timeline record for operational events **Task · Medium priority · Foundational** noCV practice brief v5 · BTRIAGE-101 · Incident assistant with read-only tools Fictional engineering practice briefs. Starter repositories, fixtures, automated grading, and verified ownership are not included. Phase: Scope incident inputs. Depends on: No preceding ticket. Difficulty: Foundational. Estimated focused work: 60 minutes; setup and prerequisite tickets are additional. Estimated field mix: Data engineering 60% · Site reliability 40%. Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence. Alerts and deployment records disagree about timestamp formats. Acceptance criteria - Normalize timestamps to UTC. - Preserve source identity and original timestamp. - Mark missing or uncertain event times. Implementation constraints - Do not invent ordering for ambiguous records. Verification - Merge valid records in stable order. - Keep equal or missing timestamps visibly uncertain. Deliverables - Timeline schema. Rollout and recovery: Import synthetic records first; retain original source records. Project prerequisites: Author synthetic alerts, deployment records, and runbooks plus deterministic tool and model doubles. Engineer value: Practice constrained tool use, operational uncertainty, and traceable AI outputs. Company value: Produce concise incident drafts that preserve source traceability and operator control. AI tools are welcome during implementation. Record assumptions, review the result, and verify its behavior. Planning status does not create Outcome Evidence or Ownership Evidence. #### BTRIAGE-102 — Limit incident retrieval to the caller's service grants **Task · High priority · Advanced** noCV practice brief v5 · BTRIAGE-102 · Incident assistant with read-only tools Fictional engineering practice briefs. Starter repositories, fixtures, automated grading, and verified ownership are not included. Phase: Scope incident inputs. Depends on: BTRIAGE-101. Difficulty: Advanced. Estimated focused work: 180 minutes; setup and prerequisite tickets are additional. Estimated field mix: Security 60% · Applied AI 40%. Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence. A related-service lookup exposes another team's restricted incident notes. Acceptance criteria - Scope every tool read by caller grants. - Recheck access for referenced records. - Return non-enumerating denial responses. Implementation constraints - Tool definitions cannot widen caller permissions. Verification - Read an authorized service timeline. - Deny a cross-service record guessed by ID. Deliverables - Authorized tool boundary. Rollout and recovery: Disable assistance when grant resolution fails. Project prerequisites: Author synthetic alerts, deployment records, and runbooks plus deterministic tool and model doubles. Engineer value: Practice constrained tool use, operational uncertainty, and traceable AI outputs. Company value: Produce concise incident drafts that preserve source traceability and operator control. AI tools are welcome during implementation. Record assumptions, review the result, and verify its behavior. Planning status does not create Outcome Evidence or Ownership Evidence. ### Build draft assistance Constrain tools, sources, and unsupported conclusions. #### BTRIAGE-103 — Bound tool arguments and returned telemetry volume **Task · High priority · Intermediate** noCV practice brief v5 · BTRIAGE-103 · Incident assistant with read-only tools Fictional engineering practice briefs. Starter repositories, fixtures, automated grading, and verified ownership are not included. Phase: Build draft assistance. Depends on: BTRIAGE-102. Difficulty: Intermediate. Estimated focused work: 120 minutes; setup and prerequisite tickets are additional. Estimated field mix: Applied AI 50% · API design 30% · Site reliability 20%. Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence. A broad query attempts to retrieve days of raw logs. Acceptance criteria - Require bounded time range and record count. - Reject unknown tool names and properties. - Truncate with explicit completeness metadata. Implementation constraints - Use aggregate synthetic records instead of secret-bearing logs. Verification - Retrieve a bounded incident window. - Reject excessive ranges and unsupported tool arguments. Deliverables - Tool request validator. Rollout and recovery: Start with conservative limits; report incomplete context rather than expanding silently. Project prerequisites: Author synthetic alerts, deployment records, and runbooks plus deterministic tool and model doubles. Engineer value: Practice constrained tool use, operational uncertainty, and traceable AI outputs. Company value: Produce concise incident drafts that preserve source traceability and operator control. AI tools are welcome during implementation. Record assumptions, review the result, and verify its behavior. Planning status does not create Outcome Evidence or Ownership Evidence. #### BTRIAGE-104 — Keep log content from issuing new tool instructions **Bug · High priority · Advanced** noCV practice brief v5 · BTRIAGE-104 · Incident assistant with read-only tools Fictional engineering practice briefs. Starter repositories, fixtures, automated grading, and verified ownership are not included. Phase: Build draft assistance. Depends on: BTRIAGE-103. Difficulty: Advanced. Estimated focused work: 180 minutes; setup and prerequisite tickets are additional. Estimated field mix: Security 50% · Applied AI 50%. Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence. An error message contains text requesting privileged configuration reads. Acceptance criteria - Treat log text as untrusted data. - Allow only declared read-only tools. - Validate every proposed tool call independently. Implementation constraints - No shell, configuration writes, or remediation tools are exposed. Verification - Summarize a benign failure record. - Inject a privileged instruction and verify no unauthorized call occurs. Deliverables - Adversarial tool-use checks. Rollout and recovery: Block provider output on policy failure and preserve sanitized diagnostics. Project prerequisites: Author synthetic alerts, deployment records, and runbooks plus deterministic tool and model doubles. Engineer value: Practice constrained tool use, operational uncertainty, and traceable AI outputs. Company value: Produce concise incident drafts that preserve source traceability and operator control. AI tools are welcome during implementation. Record assumptions, review the result, and verify its behavior. Planning status does not create Outcome Evidence or Ownership Evidence. #### BTRIAGE-105 — Require source references for incident summary statements **Task · High priority · Advanced** noCV practice brief v5 · BTRIAGE-105 · Incident assistant with read-only tools Fictional engineering practice briefs. Starter repositories, fixtures, automated grading, and verified ownership are not included. Phase: Build draft assistance. Depends on: BTRIAGE-101, BTRIAGE-103. Difficulty: Advanced. Estimated focused work: 180 minutes; setup and prerequisite tickets are additional. Estimated field mix: Applied AI 80% · Site reliability 20%. Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence. The generated summary invents a deployment that never happened. Acceptance criteria - Bind factual statements to retrieved source identities. - Reject nonexistent and unauthorized references. - Separate observations from hypotheses. Implementation constraints - Use AiEvaluationProvider and schema-validated structured responses. Verification - Accept a source-backed timeline statement. - Reject invented deployments and uncited root-cause assertions. Deliverables - Summary validator. Rollout and recovery: Render only validated drafts; return unavailable on invalid output. Project prerequisites: Author synthetic alerts, deployment records, and runbooks plus deterministic tool and model doubles. Engineer value: Practice constrained tool use, operational uncertainty, and traceable AI outputs. Company value: Produce concise incident drafts that preserve source traceability and operator control. AI tools are welcome during implementation. Record assumptions, review the result, and verify its behavior. Planning status does not create Outcome Evidence or Ownership Evidence. #### BTRIAGE-106 — Represent competing incident hypotheses explicitly **Story · Medium priority · Advanced** noCV practice brief v5 · BTRIAGE-106 · Incident assistant with read-only tools Fictional engineering practice briefs. Starter repositories, fixtures, automated grading, and verified ownership are not included. Phase: Build draft assistance. Depends on: BTRIAGE-105. Difficulty: Advanced. Estimated focused work: 150 minutes; setup and prerequisite tickets are additional. Estimated field mix: Site reliability 60% · Applied AI 40%. Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence. A latency spike follows a deployment but also overlaps a dependency outage. Acceptance criteria - List supporting and contradicting observations per hypothesis. - Avoid declaring causation from timing alone. - Identify a bounded next observation for discrimination. Implementation constraints - Recommendations remain read-only diagnostic suggestions. Verification - Represent two plausible causes. - Keep a hypothesis unresolved when discriminating telemetry is absent. Deliverables - Hypothesis draft format. Rollout and recovery: Require operator review before acting on a hypothesis. Project prerequisites: Author synthetic alerts, deployment records, and runbooks plus deterministic tool and model doubles. Engineer value: Practice constrained tool use, operational uncertainty, and traceable AI outputs. Company value: Produce concise incident drafts that preserve source traceability and operator control. AI tools are welcome during implementation. Record assumptions, review the result, and verify its behavior. Planning status does not create Outcome Evidence or Ownership Evidence. #### BTRIAGE-107 — Cancel obsolete drafts when new incident context arrives **Bug · High priority · Intermediate** noCV practice brief v5 · BTRIAGE-107 · Incident assistant with read-only tools Fictional engineering practice briefs. Starter repositories, fixtures, automated grading, and verified ownership are not included. Phase: Build draft assistance. Depends on: BTRIAGE-105. Difficulty: Intermediate. Estimated focused work: 150 minutes; setup and prerequisite tickets are additional. Estimated field mix: Applied AI 60% · Backend 40%. Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence. An older slow response overwrites the summary after a recovery event arrives. Acceptance criteria - Bind drafts to context revision. - Cancel or discard superseded work. - Preserve the latest accepted timeline revision. Implementation constraints - Record provider/model/version, timeout, retries, cost, and trace safely. Verification - Complete the newest draft first. - Resolve an older request late and reject its update. Deliverables - Revision-aware draft coordinator. Rollout and recovery: Keep the last valid draft visible with its timestamp when refresh fails. Project prerequisites: Author synthetic alerts, deployment records, and runbooks plus deterministic tool and model doubles. Engineer value: Practice constrained tool use, operational uncertainty, and traceable AI outputs. Company value: Produce concise incident drafts that preserve source traceability and operator control. AI tools are welcome during implementation. Record assumptions, review the result, and verify its behavior. Planning status does not create Outcome Evidence or Ownership Evidence. ### Assess operator usefulness Evaluate bounded behavior and safe failure handling. #### BTRIAGE-108 — Prevent repeated refreshes from exhausting incident budgets **Task · High priority · Intermediate** noCV practice brief v5 · BTRIAGE-108 · Incident assistant with read-only tools Fictional engineering practice briefs. Starter repositories, fixtures, automated grading, and verified ownership are not included. Phase: Assess operator usefulness. Depends on: BTRIAGE-107. Difficulty: Intermediate. Estimated focused work: 150 minutes; setup and prerequisite tickets are additional. Estimated field mix: Applied AI 60% · Backend 40%. Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence. Several operators refresh the same incident and trigger duplicate model calls. Acceptance criteria - Coalesce identical in-flight context requests. - Enforce per-incident reservation and timeout. - Settle usage once per provider attempt. Implementation constraints - Provider failure cannot trigger unbounded retry loops. Verification - Share one in-flight synthetic response. - Exhaust the budget and return a bounded unavailable state. Deliverables - Incident request budget. Rollout and recovery: Default to deterministic mode; pause new refreshes when accounting is uncertain. Project prerequisites: Author synthetic alerts, deployment records, and runbooks plus deterministic tool and model doubles. Engineer value: Practice constrained tool use, operational uncertainty, and traceable AI outputs. Company value: Produce concise incident drafts that preserve source traceability and operator control. AI tools are welcome during implementation. Record assumptions, review the result, and verify its behavior. Planning status does not create Outcome Evidence or Ownership Evidence. #### BTRIAGE-109 — Evaluate summary utility without rewarding confident guesses **Task · High priority · Expert** noCV practice brief v5 · BTRIAGE-109 · Incident assistant with read-only tools Fictional engineering practice briefs. Starter repositories, fixtures, automated grading, and verified ownership are not included. Phase: Assess operator usefulness. Depends on: BTRIAGE-104, BTRIAGE-106, BTRIAGE-108. Difficulty: Expert. Estimated focused work: 300 minutes; setup and prerequisite tickets are additional. Estimated field mix: Applied AI 50% · Quality engineering 30% · Site reliability 20%. Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence. The team needs a useful assessment that does not equate fluent prose with correct incident analysis. Acceptance criteria - Measure citation validity, omitted critical observations, and unsupported assertions separately. - Include partial, contradictory, and malicious context cases. - Report deterministic boundary results separately from live-model quality. Implementation constraints - Do not claim reduced incident duration from this local prototype. Verification - Detect an invented root cause. - Accept a concise unresolved summary with valid diagnostic next steps. Deliverables - Evaluation protocol and failure report. Rollout and recovery: Expand incident coverage only after reviewing unsupported-claim failures. Project prerequisites: Author synthetic alerts, deployment records, and runbooks plus deterministic tool and model doubles. Engineer value: Practice constrained tool use, operational uncertainty, and traceable AI outputs. Company value: Produce concise incident drafts that preserve source traceability and operator control. AI tools are welcome during implementation. Record assumptions, review the result, and verify its behavior. Planning status does not create Outcome Evidence or Ownership Evidence. #### BTRIAGE-110 — Write an operator handoff for stale or unavailable assistance **Chore · Low priority · Foundational** noCV practice brief v5 · BTRIAGE-110 · Incident assistant with read-only tools Fictional engineering practice briefs. Starter repositories, fixtures, automated grading, and verified ownership are not included. Phase: Assess operator usefulness. Depends on: BTRIAGE-109. Difficulty: Foundational. Estimated focused work: 60 minutes; setup and prerequisite tickets are additional. Estimated field mix: Site reliability 60% · Applied AI 40%. Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence. Operators must continue handling incidents when the assistant is down. Acceptance criteria - Show draft age and source window. - Link the original authorized records. - Document the manual timeline workflow. Implementation constraints - The assistant is never a dependency for incident response. Verification - Continue from source records during provider failure. - Verify stale drafts cannot appear current. Deliverables - Operator fallback guide. Rollout and recovery: Ship the fallback with the prototype and rehearse it locally. Project prerequisites: Author synthetic alerts, deployment records, and runbooks plus deterministic tool and model doubles. Engineer value: Practice constrained tool use, operational uncertainty, and traceable AI outputs. Company value: Produce concise incident drafts that preserve source traceability and operator control. AI tools are welcome during implementation. Record assumptions, review the result, and verify its behavior. Planning status does not create Outcome Evidence or Ownership Evidence.