Claude Skills by CheckPickerUpper

Deep code-quality and pre-merge review of a commit, PR, branch, or diff. Audit correctness, intent, naming, cohesion, types, boundaries, control flow, duplication, cost, allocation, and generalization. Check predictability: does a feature or fix create a second way to perform a capability the codebase already has? Compare existing implementations, reuse or extend sound patterns, and extract shared behavior without letting one feature own the abstraction or inventing flexibility. Every finding...

Pedantic structural audit of a commit, PR or branch, selected directory or module, or an entire codebase — hierarchy, nesting, naming, ownership, dependency boundaries, and growth readiness. Use when a change may distort repository shape, when a file or folder needs structural review, or when deciding whether to reorganize before expansion. Every finding must survive adversarial refutation and include evidence plus a concrete current→proposed change.

Justify by correctness, never by convention. Use when a design, fix, API, macro, library surface, hook message, or naming decision is defended by 'it's what the codebase does', 'the established pattern', 'the canonical lane does it', 'more ergonomic', 'less churn', 'already wired', 'smallest fix', 'compatibility', 'legacy path', or 'out of scope to change'. Prevalence and transition size are not evidence. Re-derive from correctness, name the bug the alternative causes, and be confident in the...

User-invoked modifier for evidence-first read-only investigation. Use when the user appends /readonly to a question or another skill request: answer by checking the codebase and available read-only sources, do not implement, do not open issues, do not mutate local or remote state, and do not guess with 'likely' or 'probably' when evidence can be gathered.

Review AGENTS.md changes for Markdown defects, duplicated or conflicting instructions, and unintended semantic drift. Use when editing or reviewing agent instruction files, especially large rewrites or diffs that remove existing rules.

Before implementing a new capability, service, subsystem, or integration, choose its owner—reuse, adopt, buy, build, combine, or defer—when the decision spans internal code, platform adoption, managed services, or custom boundaries. Use library-fit for package-only coverage questions.

Act as the conductor for delegated implementation: read its saved settings first and ask for any that are missing, then investigate, settle decisions, brief implementer agents, answer or escalate their questions, review what they return, and land it. Use when the user says to run as a conductor, set up conductor mode, coordinate or drive agents, hand issues to subagents or herdr panes, or direct implementation while keeping the reasoning and review. Pairs with implementer-mode, which the brie...

Make classes of invalid states and evaded constraints unreachable by type composition first. Use for bugs, bad APIs, models, schemas, generated artifacts, and enforcement rules. Treat the target language's type system as the enforcer, accept its limits, and never substitute runtime checks, brands, casts, or workarounds for a type-level construction.

Make, revise, or republish a claude.ai Artifact of any kind: report, audit, decision page, dashboard, tracker, form, tool, game, diagram, deck, or doc. Triggers: 'make an artifact', 'publish this as a page', 'turn this into an artifact', 'build a dashboard', 'update the artifact', and every revision after the reader says a page is too long, too thin, too technical, or not something they can act on. Owns the reader's job, sourced facts, information density, revision without overcorrection, and...

Adversarial stress-test of an architecture, technical proposal, API contract, state machine, or plan before implementation. Clamps the target into six orthogonal stress vises: state explosion, temporal/interruption breakdown, adversarial intent evasion, cascading failure attribution, asymmetric load, and multi-owner drift. Prohibits severity theater: every fracture must demonstrate a concrete breaking execution trace and survive an adversarial refutation pass or be destroyed. Outlaws verdict ...

Write and run focused tests without fixing implementation failures. Use when the user wants to know whether tests catch a defect and wants each result classified as an implementation bug, test-writing failure, harness failure, or pass, with the expected behavior stated.

Compose or retrieve a gist-backed agent goal from a task or GitHub issues. Use when handing work to another agent and require a human outcome, live issue references, real behavior proof, and an explicit hard-blocker contract.

Dependency trace: clarify scope, close the recursive capability graph, and derive the correctness order before implementation. Use when the user asks what a feature uses, what its dependencies use, dependency order, foundations, or a lowest-to-highest flow.

A high-judgment design interview that sharpens plans in useful batches and records settled domain language and ADRs as the conversation progresses.

Create GitHub issues from a plan, bug, finding, audit or PRD, whose titles and bodies a stranger can act on. Use when filing issues or turning work into issue-sized slices. Routes unresolved decisions to a plain-language Question/Why/Questions/Done when shape; writes each issue in its verifier's language as Outcome plus Done when; sizes issue sets to independent slices; labels by issue kind with bug priority; wires real sub-issue, blocked-by, and duplicate edges rather than prose references; ...

The Zen of CheckPickerUpper: twenty-one lines that judge designs, fixes, abstractions, types, structure, and questions, each expanded and traced to its source skill. Use when settling a design choice or an ambiguous question, choosing between options, judging whether a fix, abstraction, type, or layout is right, or when someone asks for the zen of checkpickerupper.

Implement a job handed over by a conductor: work inside the brief's scope and settled decisions, ask the conductor with options when a choice is unsettled, stop on contradicting evidence, open the PR without merging, and report per acceptance criterion with proof. Use when a brief says to work under implementer-mode, when running as a subagent or herdr-driven agent on an assigned issue, or when receiving review fixes or a rebase request for a PR you opened.

Act as a rigorous senior engineering advisor for debugging, design decisions, code and architecture questions, implementation choices, reviews, and tradeoffs. Use for questions such as how to fix a problem, whether an approach is sound, or what an experienced engineer would do; adapt to the request instead of forcing an architecture audit.

Wheel-reinvention audit: inspect implementation plans, code changes, and diffs for standard capabilities hand-rolled where the project, language, platform, framework, or an available dependency could own them. Use during implementation or review when custom parsing, validation, serialization, HTTP, retries, caching, date/time, routing, state, UI primitives, persistence, or data-structure code appears; report a replacement only with manifest and primary-source coverage evidence.

Explain a code change, diff, branch, commit, or PR as a rich, self-contained page: background, intuition with concrete examples, a walkthrough in execution order, what could break, and a quiz that cannot be gamed by picking the longest answer. Use when the user asks to explain, walk through, teach, or help them understand a change, diff, branch, or PR. Output is HTML by default, Markdown with --md (cheapest), and --chat moves the quiz into conversation as free-response questions.