Use Azure Boards through Microsoft Entra OAuth, PAT, or bearer-token auth for Work Item operations, AI Project Cockpit reporting, history/flow mining, comment intelligence, role-based delivery reports, process policy checks, and safe write-preview workflows.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Use Azure Boards through Microsoft Entra OAuth, PAT, or bearer-token auth for Work Item operations, AI Project Cockpit reporting, history/flow mining, comment intelligence, role-based delivery reports, process policy checks, and safe write-preview workflows.
Azure Boards
Use this skill when the user wants to work with Azure DevOps Boards, Work Items, project delivery status, process compliance, delivery risk, bottlenecks, comments, role-specific reporting, governance, or process policy analysis.
App Directory Review Notes
When preparing this plugin for OpenAI App Directory review, keep submission metadata conservative and evidence-backed:
Do not publish example support addresses, local repository URLs, development-only URLs, or Microsoft-owned documentation URLs as app-owned support, repository, privacy, terms, or website links.
Use deployment placeholders such as TODO_DEPLOY_HOSTED_MCP_URL or TODO_ENTRA_CLIENT_ID only in pre-submission artifacts when the real hosted MCP endpoint or Microsoft Entra client id is not known.
Do not fabricate privacy policy or terms URLs. Require publisher-owned privacy and terms pages before final submission.
Prefer individual Microsoft Entra login for app review. PAT and bearer-token modes are development or controlled automation options and must not require users to paste secrets into ChatGPT.
Confirm deployed MCP transport, OAuth registration, support contact, privacy policy, terms, screenshots, tool annotations, and output schemas before claiming App Directory readiness.
Authentication Guidance
Prefer the active configured mode reported by azure_boards_auth_status.
Use PAT mode when AZURE_BOARDS_PAT or AZURE_DEVOPS_PAT is configured.
Use existing bearer-token mode when AZURE_BOARDS_BEARER_TOKEN is configured.
Use Microsoft Entra OAuth device login when token env vars are not configured and user login is appropriate. Call azure_boards_login before data access.
Never print tokens, PATs, bearer tokens, device-code secrets, or token-cache contents.
Workflow
Check authentication with azure_boards_auth_status; call azure_boards_login only when OAuth login is needed.
Use azure_boards_query_work_items or azure_boards_get_work_item to gather board truth before making recommendations.
Prefer explainable AI tools for process and delivery insights, especially:
azure_boards_ai_delivery_risk_radar
azure_boards_ai_project_cockpit
azure_boards_ai_watchlist_report
azure_boards_ai_action_plan
azure_boards_ai_workflow_conformance
azure_boards_ai_sla_aging_monitor
azure_boards_ai_bottleneck_mining
azure_boards_ai_governance_score
azure_boards_ai_policy_gap_detector
azure_boards_ai_create_process_baseline
azure_boards_ai_process_drift_detection
azure_boards_ai_cost_of_delay_radar
azure_boards_ai_process_simulator
azure_boards_ai_capacity_forecast
azure_boards_ai_brief_export
azure_boards_ai_delivery_system_correlation
azure_boards_ai_wsjf_consistency_check
azure_boards_ai_business_value_estimate
azure_boards_ai_attachment_evidence_summary
azure_boards_ai_close_candidates
azure_boards_ai_parent_child_cleanup
azure_boards_ai_bulk_close_preview
azure_boards_ai_requirement_decision_cockpit
azure_boards_ai_evidence_first_requirement_review
azure_boards_ai_cio_requirement_risk_view
azure_boards_ai_portfolio_rationalization
azure_boards_ai_benefit_realization_tracking
azure_boards_ai_cost_avoidance_by_closure
azure_boards_ai_erp_domain_impact_scoring
azure_boards_ai_closure_governance_ledger
Safe Write-Preview Workflow
Use write-preview behavior for any create, update, comment, relation, assignment, state transition, tag, area, iteration, or policy-remediation action:
Gather the current Work Item data.
Produce a preview containing Work Item ids, fields or relations affected, old values, proposed values, and rationale.
Ask for confirmation when the requested write is not already explicit and exact.
Execute only the write-specific tool after confirmation.
Report completed changes, skipped changes, and Azure DevOps errors.
For bulk close work:
Always re-read current state, type, direct children, and parent relations before planning.
Use azure_boards_ai_bulk_close_preview before writing; it must include ids, current states, target states, child impact, comments, JSON Patch previews, skipped items, and risk.
Use azure_boards_apply_bulk_close_plan only when the preview is explicitly approved with approved:true and confirmPhrase: "APPLY_BULK_CLOSE".
Write the explanatory comment before the state transition.
After applying, re-query the relevant WIQL and report remaining counts.
Guardrails
Do not invent project status. Use Azure Boards data or clearly label assumptions.
Do not auto-assign owners, change states, or create audit evidence without user confirmation.
Treat AI scores as decision support, not as authoritative process compliance findings.
For executive summaries, include risks, evidence, and next actions grounded in Work Item fields, comments, and relations.
For comment intelligence, distinguish direct evidence from inferred intent or sentiment.
For role-based reports, avoid hiding uncertainty. Include missing data and policy/config assumptions.
For process policy checks, do not treat local defaults as organization policy unless the user or config says so.
Use azure_boards_package_health and azure_boards_auth_environment_check for diagnostics, but do not ask the user to paste secret values.
Use azure_boards_ai_plan_approved_actions only after explicit approval; it prepares batches but still does not perform Azure DevOps writes.
Use azure_boards_apply_bulk_close_plan only for a preview generated by azure_boards_ai_bulk_close_preview; never construct hidden bulk writes.
Use azure_boards_apply_test_case_plan only for a preview generated by azure_boards_ai_generate_test_cases; require approved:true and confirmPhrase: "APPLY_TEST_CASE_PLAN".
Use azure_boards_apply_traceability_plan only for a preview generated by azure_boards_ai_requirement_test_traceability; require approved:true and confirmPhrase: "APPLY_TRACEABILITY_PLAN".
Treat decision, portfolio, CIO, steering-pack, policy-as-code, evidence-ledger, and Governance Operating System tools as advisory no-write tools. They can recommend, rank, simulate, draft rewrite patches, and export evidence previews, but only explicit write tools may change Azure Boards.
Do not store PATs, bearer tokens, OAuth token caches, or live attachment extraction artifacts in plugin docs or local JSON reports.
Remember Azure DevOps process quirks: Work Item comments use 7.1-preview, and Work Item batch fetches should not combine explicit fields with relation expansion.
For AI Project Cockpit requests, combine delivery risk, SLA aging, bottleneck, governance, policy, blocker, and next-action signals into one evidence-backed report.
For history and flow mining, inspect revisions, state transitions, cycle time, reopen patterns, handoffs, stale work, and blocked periods before recommending process changes.
For comment intelligence, ground findings in Work Item comments and discussion history. Extract blockers, commitments, decisions, unresolved questions, escalations, and follow-up actions without overstating sentiment.
For role-based reports, tailor the same board truth to the requested audience:
Executives: health, risk, dates, decisions needed, and business impact.
Compliance/governance: evidence, missing fields, audit risks, and policy exceptions.
For process policy config, use explicit thresholds and rules when provided. If no config is supplied, state the default assumptions before scoring SLA, governance, transition, required-field, or evidence checks.
For proactive operations, use azure_boards_ai_watchlist_report before azure_boards_ai_action_plan; the action plan can propose patches, but does not apply them.
For baseline/drift work, create or load a baseline first, then compare current board evidence and explain which metric moved beyond threshold.
For economic prioritization, use cost-of-delay and capacity forecasts as decision support; include assumptions rather than presenting the score as finance truth.
For repo/pipeline correlation, use PR/build/pipeline/release list tools read-only and clearly separate Azure Boards evidence from delivery-system evidence.
For policy packs, validate the policy pack first and state whether findings come from user policy, example policy, or defaults.
For persistence, store only user-approved reports/baselines/watchdog snapshots in the local store; never store PATs, bearer tokens, API keys, or raw token-cache contents.
For optional LLM synthesis, use deterministic reports as the source of truth. Treat LLM output as wording/synthesis only, not as the scoring authority.
For WSJF and business-value reviews, compare Custom.BusinessValue, Custom.TimeCriticality, Custom.RiskReduction, Custom.CostOfDelay, and Custom.JobDuration against Description, attachments, status, priority, children, and aging. Treat USD-denominated values as conservative estimates and state assumptions.
For attachment-backed decisions, use azure_boards_ai_attachment_evidence_summary with extracted text when available. docx, xlsx/xlsm, pdf, xml, csv, and txt are usually useful; .msg and images are evidence metadata unless OCR/parser support is available.
For Requirement Decision Engine work, use azure_boards_ai_requirement_decision_cockpit for accelerate/review/park/close support, azure_boards_ai_evidence_first_requirement_review for missing evidence, and azure_boards_ai_cio_requirement_risk_view for executive risk. These tools support decisions; they do not justify automatic closure by themselves.
For portfolio economics, use azure_boards_ai_portfolio_rationalization, azure_boards_ai_benefit_realization_tracking, azure_boards_ai_cost_avoidance_by_closure, and azure_boards_ai_erp_domain_impact_scoring. State assumptions for USD-denominated cost, benefit, and domain-impact estimates.
For audit and governance work, use azure_boards_ai_closure_governance_ledger, azure_boards_ai_audit_decision_log, azure_boards_ai_board_hygiene_automation_preview, and azure_boards_ai_evidence_pack_completeness. Distinguish actual Azure Boards evidence from inferred governance status and never fabricate evidence ledger entries.
For outcome and CIO steering work, use azure_boards_ai_outcome_realization_cockpit, azure_boards_ai_business_case_generator, azure_boards_ai_value_leakage_detector, azure_boards_ai_board_due_diligence_report, and azure_boards_ai_steering_committee_pack. Treat generated business cases and packs as drafts requiring human review.
For traceability and process criticality, use azure_boards_ai_decision_traceability_graph, azure_boards_ai_erp_process_criticality_model, azure_boards_ai_requirement_invest_divest_matrix, azure_boards_ai_change_portfolio_simulator, and azure_boards_ai_policy_as_code_evaluation. Separate simulated closure effects from actual board state.
For Governance Operating System work, use azure_boards_ai_autonomous_board_auditor, azure_boards_ai_requirement_rewrite_studio, azure_boards_ai_decision_meeting_copilot, azure_boards_ai_cleanup_campaign_manager, azure_boards_ai_financial_backlog_ledger, azure_boards_ai_requirement_confidence_score, azure_boards_ai_dependency_blocker_graph, azure_boards_ai_process_owner_control_tower, azure_boards_ai_migration_cutover_readiness, azure_boards_ai_exception_register, azure_boards_ai_benefit_realization_followup, azure_boards_ai_operating_rhythm_planner, azure_boards_ai_okr_alignment_scorer, azure_boards_ai_compliance_readiness_review, azure_boards_ai_handover_pack_generator, and azure_boards_ai_portfolio_fitness_index. Treat rewrite suggestions, cleanup campaigns, cutover readiness, exception registers, benefit follow-ups, handover packs, strategic alignment scores, compliance readiness checks, and portfolio fitness scores as advisory previews until the user explicitly approves concrete writes.
For elicitation, conversion, test generation, traceability, mockup, diagram, SOP, prompt, and admin work, use the Copilot4DevOps gap-closure tools. Treat generated Requirements, test cases, UAT/regression suites, links, mockups, diagrams, SOPs, prompt runs, and admin configs as drafts or previews until explicitly approved.
For CIO decision assurance and autonomous Process Owner work, use azure_boards_ai_decision_memory, azure_boards_ai_recommendation_quality_score, azure_boards_ai_value_inflation_detector, azure_boards_ai_decision_court, azure_boards_ai_requirement_contract_lifecycle, azure_boards_ai_scenario_war_room, and azure_boards_ai_autonomous_governance_agent. Treat decision memory, recommendation learning, value inflation, decision court, contracts, scenarios, watchlists, agenda items, and action previews as advisory until explicitly approved.
For Enterprise Value & Trust work, use azure_boards_ai_business_digital_twin, azure_boards_ai_external_evidence_import, azure_boards_ai_event_log_process_mining, azure_boards_ai_stakeholder_influence_map, azure_boards_ai_roi_confidence_workflow, azure_boards_ai_enterprise_risk_heatmap, azure_boards_ai_policy_studio, azure_boards_ai_prompt_eval_suite, azure_boards_ai_model_risk_governance, and azure_boards_ai_adoption_cockpit. Treat external evidence, process logs, KPI links, stakeholder maps, ROI maturity, heatmaps, policies, prompt evals, model risk, and adoption metrics as decision-support evidence, not as authoritative truth without review.
For enterprise productization and data-moat work, use azure_boards_ai_connector_readiness_audit, azure_boards_ai_evidence_ingestion_pipeline, azure_boards_ai_security_privacy_review, azure_boards_ai_marketplace_submission_readiness, azure_boards_ai_org_rollout_readiness, azure_boards_ai_license_packaging_advisor, azure_boards_ai_customer_value_case_builder, azure_boards_ai_proprietary_signal_catalog, azure_boards_ai_autonomous_followup_scheduler, and azure_boards_ai_adoption_experiment_designer. Treat connector readiness, marketplace checklists, security/privacy controls, commercial packaging, signal catalogs, follow-ups, and adoption experiments as rollout planning evidence, not as deployment or scheduling actions.
For product operating system work, use azure_boards_product_snapshot_save, azure_boards_product_baseline_save, azure_boards_product_approval_queue, azure_boards_product_approval_apply_plan, azure_boards_product_approval_result_review, azure_boards_product_audit_trail, azure_boards_product_role_cockpits, azure_boards_product_admin_console, azure_boards_product_reminder_plan, azure_boards_product_decision_pack_export, azure_boards_product_decision_pack_import, azure_boards_product_outcome_proof_engine, azure_boards_product_decision_memory_learning, azure_boards_product_board_to_value_mapping, azure_boards_product_autonomous_governance_plan, azure_boards_product_compliance_evidence_score, azure_boards_product_scope_creep_radar, azure_boards_product_approval_simulation, azure_boards_product_executive_steering_room, azure_boards_product_decision_knowledge_graph, azure_boards_product_ai_readiness_prompt_governance, azure_boards_product_onboarding_wizard, and azure_boards_product_policy_pack_catalog. Treat snapshots, baselines, approval queues, apply plans, result reviews, audit trails, role cockpit configs, reminders, decision packs, outcome proof, learning loops, value maps, governance plans, compliance scores, scope creep radar, approval simulations, executive steering rooms, decision graphs, AI readiness governance, onboarding plans, and policy pack catalogs as local product artifacts or no-write previews until the user explicitly approves a concrete Azure DevOps write.
For write operations, show the proposed field changes or JSON Patch first unless the user explicitly requested the exact write.
