Polyhydra Skills / GitHub & Delivery
Define a project's lean V1 from a V0 baseline, including inherited half-built work, then create a GitHub-authoritative epic, feature-story, and QA-validation shape. Use when a project needs its first coherent release definition; when the user mentions V0, V1, a V1 checklist, validation work under an epic, inherited partial code, or post-V1 stabilization.
Drop this in — save the block below as ~/.claude/skills/v1-delivery-shaping/SKILL.md, or run:
mkdir -p ~/.claude/skills/v1-delivery-shaping cat > ~/.claude/skills/v1-delivery-shaping/SKILL.md <<'EOF' # (paste the full source block below into this file) EOF
# V1 Delivery Shaping Treat every project as V0 until its V1 definition is explicitly recorded. V0 may contain working, partial, stale, experimental, or contradictory code; it is an inventory state, not a release claim. ## Shape V1 1. Refresh the project's live GitHub issue/PR state and inspect the current codebase enough to identify working and half-built paths. 2. Define V1 as the smallest coherent, useful outcome for a named user and supported environment. State its non-goals and release boundary. 3. Classify inherited partial work: - retain in V1 when it directly supports the coherent outcome; - reshape when its user value is still sound but its boundary is wrong; - defer or retire when it is incoherent, unsafe, duplicate, or unrelated. Do not force all existing code into V1 merely because it already exists. 4. Record the V1 definition in the main epic. Create bounded feature stories directly under that epic and one linked **V1 Test Checklist / QA and Validation** task. 5. Give each feature story a compact checklist with source acceptance, exact head validation/review, and a declared quality-gate plan. Require unit tests for deterministic logic, integration tests at changed boundaries, automated execution where practical, and screenshot evidence for UI work. ## Execute V1 - Complete a feature story when its scoped source acceptance, required checks, review, and quality gates pass. Do not keep its source PR open solely for later environment-bound QA evidence. - For a website or app story, capture the intended user journey at each supported viewport, platform, or device state named by the story. Attach privacy-safe result images to the PR or QA record, then have AI compare them with the acceptance criteria, approved design, or prior baseline. Turn every concrete gap into a severity decision or linked issue; AI comparison informs the decision but does not replace functional proof, human product judgment, accessibility validation, or release authority. - Update the QA task with exact commit/version, environment, date, result, and an evidence link or precise blocker. - Keep release readiness separate from source delivery. Never call a merged PR a deployment or treat local tests as live proof. - Repair P0/P1 findings before the affected merge or release when they invalidate promised behavior. Create linked follow-up stories for P2 and lower findings unless the user explicitly keeps them in the active PR. - Preserve normal authority boundaries: do not merge, deploy, release, close issues, or change credentials without the required authorization and proof. ## V1 Definition Template ```markdown ## V1 Definition ### Coherent outcome <Who can do what, in which supported environment?> ### In scope - <existing capability to retain> - <bounded feature story> ### Explicitly out of scope - <deferred, retired, or later-version work> ### Inherited V0 work | Existing code or issue | Keep, reshape, defer, or retire | Reason | | --- | --- | --- | ### Delivery shape - Feature stories: #<issue>, #<issue> - QA and Validation task: #<issue> - Release boundary: <what evidence makes V1 release-ready> ### Quality gates - Unit tests: <deterministic behavior covered> - Integration tests: <changed boundary covered> - Automation: <CI or repeatable local command> - Visual proof: <required views, reference, and screenshot locations> ``` ## After V1 Move the completed project into stabilization by default: prefer defects, regression coverage, migration safety, observability, compatibility, performance, and documented operations. Admit new feature scope only with an explicit product reason and a new bounded story. Plan validation and rollback before changes that touch persistence, authentication, deployment, infrastructure, or external integrations.