Workflow recipes
Combine a small number of Adaptive Skills for common delivery, debugging, design, and governance tasks.
Recipes are starting patterns, not automatic pipelines. Run one skill at a time, inspect its output, and decide whether the next skill is still necessary.
Deliver a new feature
intent-clarification— confirm outcome, constraints, success, and unresolved ambiguity.feature-value-governance— check whether the feature deserves investment when value is uncertain.feature-planning— define the smallest useful slice and its proof.lean-implementation— implement only the approved slice.testing— choose and run risk-appropriate proof.handoff-summary— return verified state and the next safe step.
Stop early when: intent or value remains unresolved. Do not convert ambiguity into implementation scope.
Debug a production issue
workflow— state the incident boundary and closure proof.debugging— reproduce, isolate, fix, and guard recurrence.testing— validate the failure mode and nearby risk.communication— explain impact, resolution, and remaining uncertainty.handoff-summary— preserve verified incident context.
Stop early when: the issue cannot be reproduced. Record the evidence gap instead of guessing at a fix.
Review a product experience
ux-strategy— clarify the experience goal and plausible directions.heuristic-audit— identify usability failures in the current interface.ux-writing— improve labels, guidance, errors, and decision language.triad-check— reconcile product, design, and engineering consequences.
Optional: add ux-provocation before commitment when the team has converged too quickly on one direction.
Make an architecture decision
domain-language-alignment— establish shared terms.architecture-review— compare boundaries, coupling, complexity, and maintenance cost.premortem— identify failure conditions and warning signals.triad-check— review cross-functional consequences when the decision affects the product experience.communication— record the decision and trade-offs.
Govern a feature portfolio decision
revenue-lever-mapping— make the intended value mechanism explicit.opportunity-tree-alignment— connect the bet to an outcome and opportunity.feature-complexity-audit— estimate permanent carry cost.feature-value-governance— make the auditable investment judgment.sunset-decision— use only for an existing feature under keep-or-remove review.
Use governed knowledge safely
knowledge-source-evaluation— assess whether the source is suitable and at what maturity.restricted-context-check— review leakage, injection, permission, and contamination risks.knowledge-conflict-resolution— resolve decision-relevant disagreement between accepted sources.
Boundary: these skills do not grant access, override permissions, or turn untrusted content into authoritative instructions.
Close a long work session
checkpoint-review— decide whether to continue, stop, or hand off.qa-review— check consistency and operational risk if meaningful work changed.handoff-summary— package verified state without dragging the full session forward.
Next steps
- Use Choose the right skill to adapt a recipe to the current bottleneck.
- Read How to use a skill before invoking the first skill.
- In AletheIA, keep the Work Slice as the governing frame and use skills only inside that boundary.