agile-lifecycle

Generate a lifecycle artefact from template. Without args, lists available templates for current phase.

Evaluate and log a change request — determines if change is incremental or significant, triggers change control process if needed.

Execute a formal gate review. Checks gate criteria, validates artefacts, produces gate-review-report with PASS/FAIL/WAIVED outcome.

Initialize agile-lifecycle framework for a new project. Creates lifecycle-state.json, directory structure, and bootstraps Phase 1.

Generate a metrics report for the current phase or overall lifecycle — delivery, quality, product, AI, governance metrics.

Start or resume a lifecycle phase. Validates entry criteria, creates phase contract, invokes phase agents.

Facilitate a sprint or phase retrospective — what went well, what to improve, action items.

Add or update entries in risk register, assumption register, clarification log, or dependency log.

Plan or review the current sprint — sprint goal, committed backlog, capacity, acceptance criteria.

Show current lifecycle status — active phase, gate status, open risks, pending artefacts, and metrics summary.

Configure lifecycle tailoring for the project type — SaaS, web, desktop, CLI, AI/ML product. Adjusts gates, artefacts, and phase depth.

Use this agent to screen technical and commercial feasibility of an AI initiative. Examples: "Assess feasibility of this AI solution", "Is our data good enough for this model?", "Check technical constraints before we commit", "Go/no-go assessment for the opportunity", "Can we build this with the data we have?" <example> Context: Team wants to know if building a real-time fraud detection model is technically feasible given their infrastructure. user: "Screen the feasibility of our real-time fraud detection initiative before we commit the team" assistant: "I'll use the feasibility-screening agent to assess technical complexity, data availability, regulatory constraints, and commercial viability for this initiative." <commentary> Pre-commitment feasibility check — agent evaluates all viability dimensions before resources are allocated. </commentary> </example> <example> Context: Leadership is concerned about data privacy regulations affecting a proposed customer analytics platform. user: "What are the feasibility blockers for a GDPR-compliant customer analytics AI?" assistant: "I'll use the feasibility-screening agent to assess regulatory feasibility, data governance requirements, and technical architecture constraints for GDPR compliance." <commentary> Regulatory feasibility concern — agent evaluates compliance constraints as a first-class feasibility dimension. </commentary> </example>

Use this agent to map solution hypotheses, define value and AI hypotheses, and propose experiments. Examples: "Map our solution hypotheses", "Define value hypotheses for this AI feature", "Create the hypothesis canvas for Gate A", "What experiments should we run to test our assumptions?", "Frame our AI approach as testable hypotheses" <example> Context: Team has validated the problem and needs to structure their assumptions before designing the solution. user: "We know the problem is real — now map out all our hypotheses before we start designing" assistant: "I'll use the hypothesis-mapping agent to structure your desirability, feasibility, and viability hypotheses into a prioritized hypothesis map with validation methods." <commentary> Pre-design hypothesis structuring — agent organizes all assumptions into a testable hypothesis map before solution work begins. </commentary> </example> <example> Context: Hypothesis about user adoption of an AI recommendation feature is untested and blocking the Phase 2 go/no-go decision. user: "Our biggest assumption is that users will trust AI recommendations — how do we map and prioritize this?" assistant: "I'll use the hypothesis-mapping agent to frame this as a formal hypothesis, assess risk level, and design a lean validation experiment for Phase 1." <commentary> Critical adoption hypothesis needs formal framing — agent structures it for rigorous testing before Phase 2 commitment. </commentary> </example>

Use this agent when starting Phase 1 of the lifecycle — framing the opportunity, articulating the business problem, building the AI justification, and mapping stakeholders. Examples: "Let's frame our AI opportunity", "I need to write an opportunity statement", "Help me articulate the business case for using AI here", "Map out our stakeholders for this initiative" <example> Context: An executive sponsor has identified a potential AI initiative but needs it formally framed before board approval. user: "We want to build an AI-powered churn prediction model — help me frame the opportunity" assistant: "I'll use the opportunity-framing agent to structure this initiative with a clear problem statement, strategic alignment, and initial business case." <commentary> New AI initiative needs formal opportunity framing — this agent structures the business case and stakeholder context before any technical work begins. </commentary> </example> <example> Context: Product team has a vague idea about using ML for customer segmentation and needs to determine if it deserves investment. user: "Is there a real opportunity here for ML-based segmentation, or is this just a tech solution looking for a problem?" assistant: "I'll use the opportunity-framing agent to assess strategic fit, define the target customer, and articulate the value hypothesis before committing resources." <commentary> Opportunity validation question — agent applies structured framing to determine whether the initiative has strategic merit. </commentary> </example>

Use this agent to validate the problem with evidence through user research, pain point mapping, and market context analysis. Examples: "Validate our problem statement with user research", "Map user pain points for this problem", "What evidence do we have that this problem is real?", "Conduct a user journey analysis", "Confirm the problem is worth solving before we go further" <example> Context: Product team has framed a problem around manual data entry but hasn't confirmed users actually experience this as painful. user: "We think users hate manual data entry — how do we validate this is a real problem worth solving?" assistant: "I'll use the problem-validation agent to design a validation approach: user interviews, observation sessions, and quantitative usage data analysis to confirm problem significance." <commentary> Problem significance unclear — agent designs a validation plan before the team commits to building a solution. </commentary> </example> <example> Context: Stakeholders have conflicting views on whether the problem is customer-facing or an internal operations issue. user: "Some say it's a customer problem, others say it's operations — help us validate which one" assistant: "I'll use the problem-validation agent to structure the investigation, define the primary problem owner, and design evidence-gathering activities to resolve the conflict." <commentary> Problem ownership ambiguity — agent structures the validation to resolve the conflict with evidence rather than opinion. </commentary> </example>

Use this agent to plan iterations — break the solution into sprints, set iteration goals, define acceptance criteria, and structure the delivery roadmap. Examples: "Plan our delivery iterations", "Break down the solution into sprints", "Create our iteration plan", "How do we structure the delivery of this AI system?", "Define our sprint cadence and goals for Phase 2" <example> Context: Phase 2 design is complete and the team needs to plan delivery iterations for Phase 3-4 before committing to a Phase 2 exit. user: "Help us plan the delivery iterations — we have 6 weeks for phase 3 and 4" assistant: "I'll use the iteration-planning agent to design the sprint structure, allocate capacity, and define sprint goals across the 6-week delivery window." <commentary> Delivery planning at Phase 2 — agent structures iterations with realistic capacity and sprint goals before Phase 3 starts. </commentary> </example> <example> Context: Team velocity has been lower than expected and the Phase 4 sprint plan needs to be revised mid-delivery. user: "We're running behind — 3 sprints completed but only 60% of planned scope done. Replan the remaining iterations." assistant: "I'll use the iteration-planning agent to reassess remaining capacity, re-prioritize the backlog, and produce a revised sprint plan for the remaining iterations." <commentary> Mid-delivery replan request — agent recalculates realistic scope based on actual velocity and revised capacity. </commentary> </example>

Use this agent to build and maintain the comprehensive Phase 2 risk register covering technical, AI/data, and project risks. Examples: "Build the risk register for this phase", "Identify all risks for our AI system", "What are the data risks we should track?", "Create mitigation plans for our top risks", "Compile the full risk register for Gate B" <example> Context: Architecture review surfaced three technical risks that need to be formally logged before Gate B. user: "Add these architecture risks to the register: data latency, model drift, and vendor lock-in" assistant: "I'll use the risk-register agent to log these three technical risks with probability/impact assessment, mitigation strategies, and assigned owners." <commentary> Post-architecture review risk capture — agent formalizes risks into the register with quantified assessment and mitigation plans. </commentary> </example> <example> Context: Sponsor wants a risk summary before approving the Gate B progression. user: "Give me the risk status for Gate B — what are our top risks and are they under control?" assistant: "I'll use the risk-register agent to produce a gate-ready risk summary with top-5 risks, current mitigation status, and residual risk assessment." <commentary> Pre-gate risk review for sponsor — agent produces executive risk summary with mitigation effectiveness assessment. </commentary> </example>

Use this agent to design the solution architecture including technical design, AI/ML architecture, data architecture, and Architectural Decision Records. Examples: "Design the solution architecture for our AI product", "Create the technical architecture document", "Define our data architecture and AI pipeline", "Write ADRs for our key technical decisions", "What architecture approach should we use for this ML system?" <example> Context: Gate A approved and team is starting Phase 2 to design the technical solution for an AI recommendation engine. user: "Design the solution architecture for our recommendation system — we need the initial architecture pack" assistant: "I'll use the solution-architecture agent to produce the initial architecture pack including system design, technology stack, data pipeline, and AI model integration approach." <commentary> Phase 2 architecture design request — agent produces the structured initial architecture pack that gates Phase 3 planning. </commentary> </example> <example> Context: Technical Lead needs to validate that the proposed ML pipeline can handle the expected data volumes before committing to the design. user: "Is our proposed architecture scalable to 10M events per day — should we change the approach?" assistant: "I'll use the solution-architecture agent to assess scalability of the proposed architecture against the volume requirements and recommend design adjustments if needed." <commentary> Architecture scalability concern — agent evaluates design against non-functional requirements and recommends changes. </commentary> </example>

Use this agent to define acceptance criteria for sprint items, including BDD-style scenarios and AI model criteria. Examples: "Write acceptance criteria for this feature", "Define BDD scenarios for the user story", "Create acceptance criteria for the AI model output", "What are the acceptance criteria for this sprint?", "Write Given-When-Then scenarios for this requirement" <example> Context: Product Manager has written user stories but the acceptance criteria are vague and untestable. user: "Turn these user stories into proper acceptance criteria — they need to be testable" assistant: "I'll use the acceptance-criteria agent to structure BDD-format acceptance criteria for each story with given/when/then scenarios covering happy path, alternative paths, and negative cases." <commentary> Vague stories need testable criteria — agent applies BDD format with explicit scenarios that developers and testers can act on. </commentary> </example> <example> Context: AI recommendation feature needs criteria that include model performance thresholds, not just functional behaviour. user: "Define acceptance criteria for our recommendation feature including AI quality thresholds" assistant: "I'll use the acceptance-criteria agent to define functional BDD criteria plus AI-specific acceptance thresholds for precision, recall, and relevance scoring." <commentary> AI feature acceptance criteria require both functional and model quality dimensions — agent covers both. </commentary> </example>

Use this agent to design and plan a sprint — define the sprint goal, commit the backlog, assess team capacity, and produce a sprint plan. Examples: "Plan sprint 3", "Design our next iteration", "What should our sprint goal be?", "Commit the sprint backlog", "Help us plan this two-week sprint", "Prepare for sprint planning" <example> Context: Phase 3 is starting and the team needs to design their first sprint for the AI product build. user: "Design sprint 1 for our team — we're building a document classification model with 5 engineers" assistant: "I'll use the sprint-design agent to structure sprint 1 with goals, ceremonies, DoD, and backlog allocation appropriate for an AI/ML classification model build." <commentary> Sprint design for a new AI build — agent structures the sprint with AI-appropriate DoD including model evaluation criteria. </commentary> </example> <example> Context: Team's retrospective flagged that sprints lack clear goals and ceremonies are ineffective. user: "Our sprints feel chaotic — redesign our sprint structure to give us more focus and discipline" assistant: "I'll use the sprint-design agent to redesign the sprint structure with explicit goals, timeboxed ceremonies, and a strengthened DoD that teams can actually commit to." <commentary> Sprint dysfunction — agent redesigns sprint structure with improved ceremony design and goal-setting practice. </commentary> </example>

Use this agent to define the test strategy for a sprint — test types, coverage targets, AI model test approach, and test data requirements. Examples: "Define the test strategy for sprint 2", "What testing approach should we use for this AI feature?", "Plan our test coverage for this iteration", "Create the test plan for the sprint", "How do we test the ML model in this sprint?" <example> Context: Phase 3 planning requires a test strategy before sprints begin to ensure quality gates are achievable. user: "Define our test strategy for the AI product — we need to know what testing we'll do and when" assistant: "I'll use the test-strategy agent to design a comprehensive test strategy covering unit, integration, model evaluation, UAT, and safety testing with coverage targets per phase." <commentary> Pre-sprint test strategy definition — agent produces the testing framework that governs all Phase 3-4 quality work. </commentary> </example> <example> Context: Team has no approach for testing AI model fairness and bias, which is a compliance requirement. user: "We need a bias testing strategy for our credit scoring model — what should we test?" assistant: "I'll use the test-strategy agent to design a bias and fairness test plan with protected attribute testing, disparate impact analysis, and regulatory compliance checks." <commentary> AI fairness testing requirement — agent designs domain-specific test approach for regulatory compliance. </commentary> </example>

Use this agent to implement AI/ML components — model training, prompt engineering, experiment logging, and model cards. Examples: "Train and evaluate the model for this sprint", "Design the prompt for this LLM feature", "Log our experiment results", "Write the model card", "Implement the AI pipeline", "What experiment approach should we use?", "Evaluate model performance against our criteria" <example> Context: Team has a trained model prototype and needs to implement it within the production architecture designed in Phase 2. user: "Our model prototype is ready — guide us through implementing it in the production pipeline" assistant: "I'll use the ai-implementation agent to guide production model implementation: training pipeline, feature engineering, model registry integration, and serving layer deployment." <commentary> Model-to-production implementation — agent structures the full ML engineering implementation against the architecture design. </commentary> </example> <example> Context: Model performance is degrading in the staging environment and the team needs to diagnose and fix the issue. user: "Model accuracy dropped from 0.87 to 0.73 in staging — what's happening?" assistant: "I'll use the ai-implementation agent to diagnose the performance degradation: data distribution shift, feature pipeline issues, or model configuration problems." <commentary> Model performance issue — agent systematically diagnoses root cause in the ML pipeline before recommending fixes. </commentary> </example>

Use this agent when executing a sprint to guide feature implementation, code review, and unit testing. Examples: "Help me implement this feature against the acceptance criteria", "Review this code for quality", "What unit tests should I write for this component?", "Guide the implementation of this sprint backlog item", "How should I structure this feature code?" <example> Context: Developer is implementing a sprint backlog item and needs guidance on meeting acceptance criteria. user: "Help me implement the user notification feature — here are the acceptance criteria" assistant: "I'll use the feature-builder agent to guide implementation against the acceptance criteria, covering code structure, unit tests, and DoD verification." <commentary> Sprint execution implementation request — feature-builder guides development against acceptance criteria with quality standards. </commentary> </example> <example> Context: Code review flagged architectural violations and the developer needs to understand how to fix them. user: "The code review says I'm not following the architecture patterns — what should I change?" assistant: "I'll use the feature-builder agent to review the code against the architecture pack and provide specific refactoring guidance to align with the defined patterns." <commentary> Architecture alignment issue — feature-builder cross-references the architecture pack and guides corrective changes. </commentary> </example>

Use this agent to integrate components, test API contracts, and validate service integrations. Examples: "Integrate these services and test the API", "Run integration tests for this sprint", "Test the integration between the ML model and the API layer", "Validate our component integrations", "What integration tests should we run?", "Generate the integration test record" <example> Context: Individual features are unit-tested and ready; now integration testing must confirm the system works end-to-end. user: "All features are built — guide us through integration testing for the full system" assistant: "I'll use the integration-engineer agent to design and execute the integration test suite covering API contracts, data flows, and end-to-end scenarios." <commentary> Post-feature integration phase — agent coordinates integration testing across all system components. </commentary> </example> <example> Context: Integration test revealed a data contract mismatch between the ML serving layer and the frontend API. user: "We found a data contract mismatch between the model API and the frontend — how do we resolve this?" assistant: "I'll use the integration-engineer agent to diagnose the contract mismatch, propose a resolution approach, and update the integration test suite to prevent regression." <commentary> Integration defect — agent traces the contract violation, proposes resolution, and strengthens regression coverage. </commentary> </example>

Use when running quality assurance activities — regression testing, performance testing, defect management, and sprint quality reporting. Triggers at Subfase 4.4 or when quality gates need validation. Example: user asks "run QA on the sprint deliverables" or "check defect log". <example> Context: Phase 4 build is complete and the QA cycle must run before Gate D. user: "All features are built and integrated — run the QA cycle for Gate D" assistant: "I'll use the quality-assurance agent to execute the Phase 4 QA cycle: system testing, regression suite, security checks, and defect triage before Gate D evidence is compiled." <commentary> Pre-gate QA execution — agent coordinates the full QA cycle with systematic defect tracking and gate evidence production. </commentary> </example> <example> Context: A critical security vulnerability was found during QA that could block Gate D. user: "QA found a SQL injection vulnerability in the admin API — how does this affect Gate D?" assistant: "I'll use the quality-assurance agent to assess the vulnerability severity, determine gate impact, and define the remediation conditions required before Gate D can proceed." <commentary> Critical defect blocking gate — agent assesses severity, determines gate impact, and defines remediation path. </commentary> </example>

Use when validating AI/ML model performance, bias, fairness, safety, or conducting red-team evaluation. Triggers at Subfase 5.2 or when AI/ML components need formal validation before release. Example: user asks "validate the AI model" or "run red-team evaluation". <example> Context: Model has been trained and integrated; Phase 5 requires formal model validation before production approval. user: "Validate the recommendation model against our Phase 1 success criteria before Gate E" assistant: "I'll use the ai-model-validation agent to evaluate the model against all defined performance thresholds, run bias tests, and compile the model validation report for Gate E." <commentary> Pre-gate model validation — agent executes the full validation suite against Phase 1 criteria and produces gate evidence. </commentary> </example> <example> Context: Fairness audit required by compliance before a credit scoring model can go to production. user: "Run a fairness audit on the credit scoring model — we need evidence for the compliance review" assistant: "I'll use the ai-model-validation agent to run disparate impact analysis, protected attribute testing, and calibration checks across demographic segments." <commentary> Compliance-required fairness validation — agent runs structured bias and fairness tests with documented evidence. </commentary> </example>

Use when executing end-to-end functional validation, preparing UAT, or producing the Functional Test Report. Triggers at Subfase 5.1 or when functional completeness needs to be assessed. Example: user asks "run functional validation" or "prepare UAT". <example> Context: Phase 5 starts and the built system needs user acceptance testing with real users before Gate E. user: "Phase 4 is complete — set up and run UAT for the recommendation system" assistant: "I'll use the functional-validation agent to design the UAT plan, recruit participants, run structured test sessions, and compile acceptance evidence for Gate E." <commentary> UAT coordination for Phase 5 — agent structures and executes user acceptance testing with evidence collection for the gate. </commentary> </example> <example> Context: Users are rejecting the AI recommendations in UAT but it's unclear if the issue is the model or the UI. user: "UAT participants are unhappy with recommendations — but we don't know if it's the model or the interface" assistant: "I'll use the functional-validation agent to diagnose whether the UAT failure is a model accuracy issue, a UX/explanation issue, or a user expectation issue — then recommend the correct remediation." <commentary> UAT failure root cause analysis — agent separates model quality issues from UX issues to direct the right fix. </commentary> </example>

Use when preparing a formal gate review package — collecting evidence, checking gate criteria completeness, and producing the gate pack artefact. Triggers at Subfase 5.3 or before any formal Gate review (A-F). Example: user asks "prepare Gate E package" or "check gate readiness". <example> Context: All Phase 5 activities are complete and the team needs to prepare the Gate E evidence package. user: "Phase 5 is done — help us prepare for Gate E review" assistant: "I'll use the gate-preparation agent to compile all required evidence for Gate E, verify completeness against the gate criteria, and structure the gate review package." <commentary> Pre-gate evidence compilation — agent assembles the evidence package and verifies completeness before the gate review meeting. </commentary> </example> <example> Context: Gate reviewer flagged missing evidence from the previous gate attempt and the team needs to know what to fix. user: "Gate D was rejected due to missing evidence — what exactly do we need to provide?" assistant: "I'll use the gate-preparation agent to identify all missing evidence items, assign owners, and produce a remediation plan for the gate resubmission." <commentary> Gate remediation after rejection — agent diagnoses the evidence gaps and creates a structured remediation plan. </commentary> </example>

Use when facilitating stakeholder reviews, collecting sign-offs, or managing stakeholder acceptance before a gate or release. Triggers at Subfase 5.4 or when formal stakeholder approval is required. Example: user asks "coordinate stakeholder sign-off" or "run the release review". <example> Context: Phase 5 includes a formal stakeholder review session before Gate E and the team needs to run it effectively. user: "Set up and facilitate the stakeholder review for Phase 5 — sponsor and key users are attending" assistant: "I'll use the stakeholder-review agent to design the review agenda, prepare demo scenarios, and facilitate structured feedback collection from sponsor and users." <commentary> Stakeholder review facilitation — agent structures the review session to maximize feedback quality and capture actionable decisions. </commentary> </example> <example> Context: Stakeholder review resulted in contradictory feedback from different stakeholders and the team needs to resolve conflicts. user: "The sponsor wants feature X but the users don't want it — how do we handle this?" assistant: "I'll use the stakeholder-review agent to synthesize the conflicting feedback, frame the trade-offs, and facilitate a decision with the sponsor using structured options." <commentary> Conflicting stakeholder feedback — agent structures the conflict as a decision rather than leaving it unresolved. </commentary> </example>

Use when executing a production deployment — running deployment steps, recording outcomes, monitoring go-live, and deciding on rollback. Triggers at Subfase 6.2 or when a deployment needs to be executed or documented. Example: user asks "execute deployment" or "create deployment record" <example> Context: Release is approved and the deployment engineer needs to execute the production deployment. user: "Release is approved — guide us through the production deployment of the AI service" assistant: "I'll use the deployment-engineer agent to execute the deployment runbook: infrastructure provisioning, model deployment, traffic routing, and post-deployment smoke tests." <commentary> Production deployment execution — agent follows the deployment runbook and verifies each step before proceeding. </commentary> </example> <example> Context: Canary deployment detected elevated error rates and the team needs to decide whether to proceed or roll back. user: "Canary deployment shows 2% error rate vs 0.1% baseline — proceed or roll back?" assistant: "I'll use the deployment-engineer agent to assess the error rate against the rollback trigger threshold, diagnose the error pattern, and recommend proceed/rollback with evidence." <commentary> Canary anomaly during deployment — agent evaluates against pre-defined thresholds and makes a data-driven recommendation. </commentary> </example> <example> Context: Deployment completed successfully but the infrastructure cost is higher than estimated. user: "Deployment is live but infra costs are 40% over budget — what happened?" assistant: "I'll use the deployment-engineer agent to analyse the cost overrun, identify the resource configuration mismatch, and recommend optimization actions." <commentary> Post-deployment cost analysis — agent traces cost overrun to specific infrastructure choices and recommends corrections. </commentary> </example>

Use when managing the post-launch hypercare period — intensive monitoring, rapid incident response, user feedback triage, and hypercare closure. Triggers at Subfase 6.3 or when go-live is confirmed and hypercare begins. Example: user asks "start hypercare" or "generate hypercare report" <example> Context: Product launched 2 days ago and the hypercare period is active with the team monitoring for issues. user: "Day 2 of hypercare — what should we be monitoring and what counts as an incident?" assistant: "I'll use the hypercare-lead agent to define the monitoring protocol, set incident thresholds, and establish the escalation path for the hypercare period." <commentary> Hypercare setup — agent defines what to monitor, incident criteria, and escalation procedures for the post-launch period. </commentary> </example> <example> Context: User complaints about slow AI recommendations are coming in via support tickets during the hypercare period. user: "Multiple support tickets about slow recommendations — is this an incident?" assistant: "I'll use the hypercare-lead agent to assess the performance data against incident thresholds, classify the severity, and trigger the appropriate response protocol." <commentary> Potential incident during hypercare — agent classifies severity and triggers response protocol based on pre-defined thresholds. </commentary> </example> <example> Context: Hypercare period is ending and the team needs to determine if the product is ready for steady-state operations. user: "3 weeks since launch — are we ready to exit hypercare?" assistant: "I'll use the hypercare-lead agent to evaluate hypercare exit criteria: incident frequency, user adoption, system stability, and handover readiness for the operations team." <commentary> Hypercare exit assessment — agent evaluates exit criteria before transitioning to Phase 7 operations. </commentary> </example>

Use when planning a product release — release plan, go-live criteria, communication plan, rollout strategy. Triggers at Subfase 6.1 or when release planning needs to begin. Example: user asks "plan the release" or "create release plan" <example> Context: Gate E approved and the team is planning the production release of the AI product. user: "Gate E is approved — help us plan the production release for the recommendation system" assistant: "I'll use the release-manager agent to create the release plan: go-live checklist, rollback procedure, communications plan, and release timeline." <commentary> Post-gate release planning — agent structures all release activities needed for a safe production go-live. </commentary> </example> <example> Context: Release is scheduled for next week but a last-minute bug was found that may require a rollback plan. user: "We found a bug two days before release — do we proceed or delay? And what's the rollback plan?" assistant: "I'll use the release-manager agent to assess the bug severity, evaluate go/no-go criteria, and produce a rollback plan if we proceed with the release." <commentary> Last-minute release decision — agent evaluates go/no-go criteria and ensures rollback readiness regardless of decision. </commentary> </example> <example> Context: Release was successful but the communications team needs a structured post-release announcement. user: "Release went live successfully — draft the internal and external communications" assistant: "I'll use the release-manager agent to draft release communications for internal stakeholders and external users, referencing the release notes and key outcomes." <commentary> Post-release communication — agent produces stakeholder communications aligned to the release outcomes. </commentary> </example>

Use when monitoring AI/ML model performance in production — drift detection, retraining decisions, AI monitoring reports, or model version management. Triggers at Subfase 7.2 or when AI model health needs assessment. Example: user asks "check for model drift" or "generate AI monitoring report" <example> Context: ML model has been in production for 2 months and the team needs to check for performance degradation. user: "The recommendation model has been live for 2 months — check for model drift or performance issues" assistant: "I'll use the ai-ops-analyst agent to analyse model performance metrics, data distribution drift, and feature importance stability since the last evaluation." <commentary> Routine model health check — agent assesses drift and performance against the baseline established at Gate E validation. </commentary> </example> <example> Context: User feedback suggests recommendation quality has declined but the error rate is still normal. user: "Users are complaining about worse recommendations but our error rate is fine — what's happening?" assistant: "I'll use the ai-ops-analyst agent to investigate recommendation quality metrics, concept drift indicators, and user feedback patterns that may not be captured in error rate monitoring." <commentary> Quality degradation not captured by standard monitoring — agent investigates AI-specific indicators beyond operational metrics. </commentary> </example>

Use when driving continuous improvement cycles — retrospectives, improvement backlog management, change recommendations, or iteration decisions. Triggers at Subfase 7.3 or when improvement opportunities need capturing and prioritizing. Example: user asks "run a retrospective" or "prioritize the improvement backlog" <example> Context: End of Q1 with the product in production; team needs to run a lifecycle retrospective and identify improvement priorities. user: "Run a Q1 retrospective for the AI product — what should we improve?" assistant: "I'll use the continuous-improvement agent to facilitate a structured retrospective across delivery, product, and operational dimensions and produce a prioritized improvement backlog." <commentary> Quarterly lifecycle retrospective — agent structures the review and converts insights into an actionable improvement backlog. </commentary> </example> <example> Context: Post-incident review identified process gaps that need to be addressed to prevent recurrence. user: "The P1 incident revealed gaps in our incident response process — how do we improve it?" assistant: "I'll use the continuous-improvement agent to analyse the incident findings, identify systemic process gaps, and design specific improvements with owners and success metrics." <commentary> Post-incident process improvement — agent converts incident learnings into structured process changes. </commentary> </example>

Use when closing a lifecycle iteration — capturing lessons learned, archiving artefacts, and deciding whether to retire, sustain, or start a new iteration. Triggers at Subfase 7.4 or when a lifecycle iteration reaches its natural closure point. Example: user asks "close this lifecycle iteration" or "produce lessons learned" <example> Context: All Phase 7 gates are met and the product is stable; the team needs to formally close the lifecycle. user: "The product is stable and all criteria are met — how do we formally close the lifecycle?" assistant: "I'll use the lifecycle-close agent to execute the formal closure process: final gate review, knowledge capture, artefact archiving, and lifecycle state transition to CLOSED." <commentary> Lifecycle closure — agent executes the formal closure sequence and ensures all governance requirements are met before closing. </commentary> </example> <example> Context: Organisation wants to extract learnings from a completed lifecycle to improve future AI initiatives. user: "What lessons should we capture from this lifecycle for our next AI project?" assistant: "I'll use the lifecycle-close agent to structure the lessons-learned capture across all 7 phases and produce a reusable knowledge artefact for future initiatives." <commentary> Lifecycle knowledge extraction — agent structures lessons-learned capture systematically across all phases. </commentary> </example>

Use when monitoring production operations — SLA tracking, incident management, service reporting, or capacity planning. Triggers at Subfase 7.1 or when service health needs assessment. Example: user asks "generate service report" or "check SLA compliance" <example> Context: Product is in Phase 7 steady-state operations and the team needs to review operational health. user: "Give us the operational health summary for the AI recommendation service" assistant: "I'll use the operations-monitor agent to review current operational metrics: availability, error rates, latency, throughput, and model performance indicators." <commentary> Operations health review — agent provides structured assessment across all operational dimensions. </commentary> </example> <example> Context: Alert fired for elevated API error rate and the on-call engineer needs guidance on response. user: "API error rate alert fired at 3am — what do I check first?" assistant: "I'll use the operations-monitor agent to guide the incident triage: check the alert context, identify affected components, and follow the runbook for API error rate incidents." <commentary> Incident response guidance — agent provides structured triage steps based on the alert type and system context. </commentary> </example>

Use when planning product retirement — retirement decision record, impact assessment, sunset plan, migration plan, and decommissioning. Triggers at Subfase 7.5 when retirement decision is approved. Example: user asks "plan product retirement" or "create sunset plan" <example> Context: Business decision made to retire the AI recommendation system and replace it with a new approach. user: "We've decided to retire the recommendation model — help us plan the decommissioning" assistant: "I'll use the retirement-planner agent to design the retirement plan: data preservation, user migration, model decommissioning, and infrastructure tear-down sequence." <commentary> Product retirement planning — agent structures the full decommissioning process with data preservation and user impact management. </commentary> </example> <example> Context: Legal team requires data retention policy compliance during the system retirement. user: "What data do we need to retain and for how long when we retire this AI system?" assistant: "I'll use the retirement-planner agent to define the data retention obligations, identify regulatory requirements, and design the compliant archiving and deletion plan." <commentary> Regulatory data retention during retirement — agent identifies compliance obligations and designs a compliant data disposal plan. </commentary> </example>

Use this agent when the user asks to generate, create, or fill an artefact or template. Examples: "Generate an opportunity statement", "Create the risk register entry for this finding", "Fill in the gate review template", "I need an assumption register entry", "Create the sprint iteration plan". <example> Context: Team is starting Phase 1 and needs the initial opportunity framing artefact. user: "Generate the opportunity framing document for our new AI recommendation system" assistant: "I'll use the artefact-generator to create the opportunity framing artefact from the Phase 1 template, populated with your initiative context." <commentary> Artefact creation request with clear phase context — artefact-generator selects the correct template and produces a populated document. </commentary> </example> <example> Context: Gate A review requires a business case document that hasn't been produced yet. user: "We need the business case artefact for Gate A — can you generate it from our discovery notes?" assistant: "I'll use the artefact-generator to produce the business case artefact using the Phase 1 template, incorporating your discovery notes." <commentary> Artefact required for gate evidence — artefact-generator maps the gate requirement to the correct template and generates the document. </commentary> </example>

Use this agent when the user asks to run a gate review, formal approval checkpoint, or quality gate. Examples: "Run gate A review", "We need to pass Gate D before release", "Perform the formal gate check for phase 2", "Is our evidence sufficient for Gate B?" <example> Context: Phase 2 work is complete and the team needs to pass Gate B before starting Phase 3. user: "We've completed all Phase 2 deliverables, can you run Gate B review?" assistant: "I'll use the gate-reviewer agent to assess Gate B evidence against the gate criteria and produce a formal gate review report." <commentary> Gate review request with specific gate — gate-reviewer assesses evidence and produces pass/fail determination with rationale. </commentary> </example> <example> Context: Sponsor wants to know if Gate D requirements have been met before approving release planning. user: "Check if we're ready for Gate D — I need to brief the sponsor" assistant: "I'll use the gate-reviewer to evaluate Gate D readiness, checking all required evidence against the gate criteria reference." <commentary> Pre-gate readiness check for stakeholder communication — gate-reviewer provides evidence status and gap analysis. </commentary> </example>

Use this agent when the user asks about lifecycle status, current phase, navigation between phases, or needs routing to the appropriate phase agent. Examples: "What phase are we in?", "Where are we in the lifecycle?", "How do I start phase 3?", "What should I do next?" <example> Context: Team is returning to work on a product initiative and wants to know where they left off. user: "What phase are we in and what should we do next?" assistant: "I'll use the lifecycle-orchestrator agent to read your lifecycle state and provide current status with next steps." <commentary> The user needs lifecycle orientation — the orchestrator reads lifecycle-state.json and routes to the right phase agent. </commentary> </example> <example> Context: Product Manager wants to start Phase 3 after completing Gate B. user: "Gate B was approved, how do I start Phase 3?" assistant: "I'll use the lifecycle-orchestrator to confirm the gate outcome, update lifecycle state, and route you to the Phase 3 sprint-design agent." <commentary> Phase transition request requires state update and routing — exactly what the orchestrator handles. </commentary> </example>

Use this agent when the user asks for metrics, reports, performance analysis, or KPI tracking. Examples: "Generate a metrics report", "What's our velocity this sprint?", "Show me AI model performance metrics", "How are we tracking against quality targets?", "Generate the governance metrics report", "Compare our delivery performance against baseline" <example> Context: End of sprint 3 and the delivery lead wants a metrics health check before Gate C. user: "Give me a metrics summary for sprint 3 — are we on track for Gate C?" assistant: "I'll use the metrics-analyst to assess sprint 3 delivery metrics, quality indicators, and Gate C readiness based on current lifecycle data." <commentary> Sprint-end metrics review with gate readiness angle — analyst reads lifecycle state and produces health dashboard. </commentary> </example> <example> Context: Post-launch review requires analysis of AI model performance against baseline KPIs. user: "How is the model performing against our Phase 1 success criteria?" assistant: "I'll use the metrics-analyst to compare current AI/ML performance metrics against the success criteria defined in the Phase 1 business case." <commentary> Post-launch AI performance review — analyst compares actuals against baseline and flags deviations. </commentary> </example>

Use this agent when the user asks to add or update risks, assumptions, clarifications, or dependencies. Examples: "Log this risk", "Add an assumption about data quality", "Update the status of risk R-003", "What assumptions are past due?", "Show me all open clarifications", "Track this dependency on the data team" <example> Context: Team is in Phase 2 and has identified that a key data source may not be accessible. user: "We might not have access to the customer transaction data — how do we track this?" assistant: "I'll use the risk-assumption-tracker to log this as a blocking assumption, assess its impact on the Phase 2 timeline, and set a resolution deadline." <commentary> New blocking assumption identified — tracker logs it with impact rating, owner, and resolution plan. </commentary> </example> <example> Context: Sprint retrospective revealed three new technical risks that need to be assessed. user: "Add these three risks from our retro to the risk register and prioritize them" assistant: "I'll use the risk-assumption-tracker to log and prioritize the three risks, updating the risk register with impact/likelihood ratings and mitigation plans." <commentary> Post-retro risk capture — tracker logs multiple risks and applies prioritization framework. </commentary> </example>

This skill should be used when a project includes AI/ML components and needs guidance on experiment design, model card creation, AI validation, bias and fairness assessment, LLM red-teaming, or drift monitoring configuration. Applies when a user asks to design an AI experiment, plan model validation, set up ML monitoring, or assess AI-specific gate criteria (Gates D and E).

This skill should be used when a user needs to create, fill, or validate a lifecycle artefact using the framework's template library — including selecting the correct template for a phase or gate artefact, filling all mandatory fields, removing guidance comments, and validating against a JSON schema. Triggers when the user says "create the [artefact name]", "fill in the template for", or "is this artefact gate-ready".

This skill should be used when shaping the product backlog, prioritizing stories, or preparing items for sprint readiness. Triggers when entering Phase 3 (Discovery), transitioning to Phase 4 (Delivery), or when backlog health needs assessment.

This skill should be used when evaluating if a change is incremental or significant, when processing a change request, or when updating the change log. Triggers when scope, requirements, or architecture changes are proposed.

This skill should be used when defining, applying, or updating the Definition of Done. Triggers when Phase 4 is about to start, when a story is declared complete, or when quality standards need to be updated across sprint and release levels.

This skill should be used when assembling gate evidence packs, tracking artefact completeness, or managing waivers before any gate review (A–F). Triggers when a phase is approaching its gate or when evidence index needs to be updated.

This skill should be used when conducting a formal gate review (Gates A-F), when validating gate criteria, or when deciding PASS/FAIL/WAIVED for a lifecycle gate.

This skill should be used when a project needs the lifecycle framework adapted to its specific product type (MVP, startup, data platform, enterprise, or regulated product), team size, or risk profile. Triggers when the user asks "what phases are mandatory for us", "can we skip Gate B", "we're building an MVP — what's the minimum governance", or when Phase 1 is starting and the governance configuration must be decided.

This skill should be used when sprint metrics, quality metrics, AI/ML model metrics, or governance metrics need to be collected, calculated against thresholds, or reported. Triggers when a user asks for a sprint metrics report, wants to check if velocity is healthy, needs an AI monitoring report, or detects a threshold breach that requires escalation. Also applies when baselining metrics at the start of Phase 4.

This skill should be used when preparing for production deployment, assessing operational readiness, or planning the ops transition. Triggers before Gate E (release readiness) or when transitioning to Phase 6.

This skill should be used when tracking deviations from signed phase contracts, managing scope changes that affect contracts, or escalating contract violations. Triggers when a change request arrives, when a phase is running beyond its contract bounds, or when a gate review reveals contractual non-compliance.

This skill should be used when creating a new phase contract at the start of a phase, validating entry criteria before phase work begins, tracking exit criteria progress during a phase, or triggering a gate review when all exit criteria are met. This skill creates and maintains phase contracts; for tracking deviations from an already-signed contract, use phase-contract-enforcement instead.

This skill should be used when creating RACI charts, mapping accountability by activity, resolving accountability gaps, or producing the Role-Responsibility Map artefact. Triggers when entering Phase 2 (Architecture) or Phase 3 (Discovery), or when team structure changes during delivery.

This skill should be used when a full sprint or phase retrospective needs facilitation, when the improvement backlog needs management, or when retrospective action items from a prior sprint need follow-up. This is the primary retrospective skill — sprint-facilitation references it for deep retrospective guidance. Triggers when a user asks for a retrospective format, wants to track improvement actions, or is running a phase retrospective as a gate artefact.

This skill should be used when managing risks, assumptions, clarifications, or dependencies. Triggers when adding new risks, updating risk status, identifying assumptions, or tracking open decisions.

This skill should be used when facilitating sprint ceremonies — sprint planning, daily standup, or sprint review. Triggers when a new sprint starts, when a standup is losing focus, or when a sprint review needs stakeholder demo structure. For deep retrospective facilitation or improvement backlog management, use the retrospective skill instead. Also applies when AI/ML experiment checkpoints need to be integrated into sprint ceremonies.

3 hooks across 2 events