bsafronov
- 20 skills
- 0 followers
- 6 hours ago last updated
- ▌ Fix Bug · bsafronov bundleRepair an established bug mechanism and prove the failing behavior is corrected without breaking affected behavior. Use when the cause and expected behavior are known. Do not use for diagnosis alone, new feature behavior, structural refactoring, or staged compatibility migration.
- ▌ Split Work · bsafronov bundleDecompose agreed scope into bounded, verifiable work items or ticket drafts with genuine prerequisites. Use when the user asks to break a plan, spec, or discussion into implementable tasks. Do not use for specification writing alone, implementation, or status and publication operations on an unchanged breakdown.
- ▌ Write Spec · bsafronov bundleSynthesize or revise a feature specification from agreed discussion, supplied requirements, and relevant repository evidence. Use when the user asks to capture intended behavior and acceptance criteria as a spec. Do not use for brainstorming alone, ticket decomposition, publishing an unchanged document, or reviewing implementation against an existing spec.
- ▌ Write Skill · bsafronov bundleCreate or revise one atomic agent Skill with a precise selection boundary and observable completion. Use when recurring work needs reusable agent judgment. Do not use for ordinary documentation or one-off instructions.
- ▌ Model Domain · bsafronov bundleEstablish or revise domain concepts, their relationships, and shared terminology from product evidence. Use when meanings conflict or a domain model needs defining, including its glossary and consequential decisions. Do not use for merely reading terminology, editorial document changes, or implementation-only architecture decisions.
- ▌ Improve Skill · bsafronov bundleImprove one agent Skill from concrete task or evaluation evidence after explicit invocation. Use when a recurring behavior gap is known between product tasks. Do not use during product delivery or for speculative cleanup.
- ▌ Map Decisions · bsafronov bundleCreate or revise a resumable map of unresolved decisions and their evidence dependencies for an effort spanning multiple tasks. Use when uncertainty prevents a reliable route to an agreed destination, or new evidence changes that route. Do not use for a single bounded decision, specification writing, decomposing settled scope, critique alone, or task creation and status operations on an unchanged plan.
- ▌ Refactor Code · bsafronov bundleImplement an agreed structural code change while preserving observable behavior. Use when extraction, consolidation, or responsibility moves must keep existing behavior equivalent. Do not use for an unresolved interface decision alone, bug correction, new feature behavior, or a staged compatibility transition.
- ▌ Challenge Plan · bsafronov bundleStress-test a user-selected plan, decision, or idea against consequential assumptions and trade-offs. Use when the user asks for critical questioning, grilling, or a plan critique. Do not use for ordinary planning or implementation, factual explanation alone, or defect-only code review.
- ▌ Evaluate Skill · bsafronov bundleEvaluate whether one agent Skill selects correctly, respects its boundary, composes cleanly, and achieves its outcome. Use when Skill behavior needs evaluation. Do not use to edit Skills or review product code.
- ▌ Migrate System · bsafronov bundleImplement and prove a requested stage of a compatibility or persisted-data transition with a viable recovery path. Use when old and new readers, writers, data forms, or deployments must coexist or recover from partial migration. Do not use for an in-place refactor, an ordinary feature or dependency update without a transition window, or ticket decomposition alone.
- ▌ Build Prototype · bsafronov bundleBuild a disposable artifact that answers a design question through concrete behavior or comparison. Use when a proposed interaction, state model, or interface needs trying before a design decision. Do not use for production implementation, visual polish alone, source research, or recording an already agreed spec or domain model.
- ▌ Investigate Bug · bsafronov bundleIdentify the mechanism behind an unexplained failure by discriminating plausible causes. Use when a bug, intermittent failure, or performance regression has an unknown cause. Do not use for repairing an established cause, implementing expected new behavior, or verifying a known change alone.
- ▌ Design Interface · bsafronov bundleDesign a code interface and its responsibility boundary from concrete caller needs. Use when choosing what a module exposes, where coordination belongs, or how callers control dependencies and lifecycle. Do not use for visual UI design, domain terminology alone, critique alone, or implementing a settled feature or refactor without an unresolved interface decision.
- ▌ Reframe Explanation · bsafronov bundleRe-explain a previous answer so the missing connection becomes concrete. Use when the user says an explanation did not land or asks to try explaining it another way. Do not use for a first explanation, shortening or translating understood text, or disagreement that asks for a new decision rather than clarification.
- ▌ Review Requirements · bsafronov bundleReview a concrete code change against its originating requirements and report missing, partial, incorrect, or unrequested behavior. Use when checking a PR, branch, commit, or WIP diff with an available request or spec. Do not use for defect-only code review, requirements drafting, or implementation.
- ▌ Synthesize Evidence · bsafronov bundleResolve a bounded research question into findings supported by attributable evidence, including applicability limits and unresolved conflicts. Use when sources must be compared, reconciled, or assessed for coverage before an answer is reliable. Do not use for a single factual lookup, locating files or links, domain decisions, or writing a specification from settled facts.
- ▌ Implement Test First · bsafronov bundleImplement requested behavior through a failing behavioral test, a minimal change, and verified passing tests. Use when the user asks for test-first development, TDD, or red-green-refactor. Do not use for ordinary implementation without that sequencing request, adding tests for existing behavior alone, or verification-only work.
- ▌ Write Agent Guidance · bsafronov bundleWrite repository steering and referenced guidance that agents can reach and apply at the right task boundary. Use when creating or restructuring agent-facing instructions, repairing missed context, or clarifying instruction completion. Do not use for Skill authoring alone, ordinary prose edits, domain terminology decisions, or merely following existing guidance.
- ▌ Write Commit Message · bsafronov bundleWrite a concise Conventional Commit message grounded in the intended change. Use when drafting a message for a diff or a commit task. Do not use for release notes, PR descriptions, or Git operations alone.