All authors
Claude Skills by KyleMit
github.com/KyleMit60 skills0 installs16 views
- Skills GuideGrouped catalog of every Splotch skill — what each one does and how related skills chain together (the audit lifecycle, the PR flow, ADRs, handoffs). Use when asked what skills are available, which skill fits a task, or how skills relate — and to register a skill you just created, renamed, or deleted.Votes: 0GitHub stars: 7
- Start Capture SessionTake over the physical iPad and Android capture rig at the start of a performance session — reserve and prove both devices, route around foreign host processes, and load the traps that produce plausible wrong numbers. Use before any physical-device profiling, targeted empirical review check, capture campaign, or performance matrix run, and when picking up devices a previous session was using.Votes: 0GitHub stars: 7
- TestingFull testing guide — the three-tier strategy (Vitest unit, Playwright E2E, Maestro native smoke on Android + iOS), every test command, CI workflow triggers, and Maestro installation. Use before writing, running, or modifying tests, or when debugging CI test failures.Votes: 0GitHub stars: 7
- Vectorize ImageTrace a bitmap (PNG/JPG/WebP/GIF/BMP) to clean vector art — SVG, PDF, EPS, DXF, or PNG — through the Vectorizer.AI API, with the credit budget guarded by a free test mode. Use when asked to vectorize, trace, or convert a raster image to SVG/vector, to turn a coloring outline or logo into scalable paths, or to call Vectorizer.AI at all. Covers auth, every API parameter, output options, errors, and the credit-cost model.Votes: 0GitHub stars: 7
- Vet AuditsAdversarially validate each finding in docs/AUDIT.md against the current code — is it real, worth solving, and actionable — then file the survivors as type:audit GitHub issues and drop the rest. Use when asked to vet, review, or prune the audit backlog before fixing it. Review only; it does not implement fixes.Votes: 0GitHub stars: 7
- Walk Through DecisionWalk through a decision in plain language — what is actually being decided, what is at stake, the real options with honest pros and cons, a table of what stays the same and what differs, and one recommended approach — tagging each decision as the agent's to make or the user's. `mode=autonomous` makes the agent's own calls instead of asking, each locked after a rival agent reviews it — or, on a low-stakes call when no rival can run, locked and marked unreviewed; a still-split high-stakes call ...Votes: 0GitHub stars: 7
- Analyze Session TranscriptsMine local Claude Code session transcripts (~/.claude/projects/ JSONL) into factual, evidence-anchored per-session reports — wrong first moves, self-corrections, failures, one-off scripts, empirical discoveries. Use when the user asks to analyze, debrief, or report on past sessions. Claude transcripts only; Codex sessions have a different shape and are out of scope.Votes: 0GitHub stars: 7
- Burn Down AuditsDrive the scripted bulk burndown of docs/AUDIT.md — one one-shot `claude -p` subprocess per role per finding (verify → implement → adversarial review → fix), orchestrated by tools/audit-burndown/ and built to run unattended. Use when the staged audit backlog is too large to vet-and-file as GitHub issues (hundreds of findings) and the user asks to burn it down in bulk, run the audit burndown, or launch/check on a run.Votes: 0GitHub stars: 7
- Reconcile Agent MemoriesReconcile a Claude Code project's persistent MEMORY.md index and memory files with the current repository and GitHub state, promoting durable lessons into shared repo guidance. Use for a periodic memory cleanup or after skill, file, script, or workflow renames; this is not a transcript analysis.Votes: 0GitHub stars: 7
- Run Rival AgentPair this session, as the native handler, with a rival agent — a fresh Codex CLI process on the ChatGPT plan's included usage — for an independent review of a PR, a diff, or a free-form question about this checkout. The rival reads a disposable worktree and asks you to run commands through a broker; you run each one under your own permissions or decline; its findings post to the PR verbatim. Use when the user asks for an outside/independent/second-opinion review, wants Codex to check the work...Votes: 0GitHub stars: 7