gnibu
UsermacOS menu bar app that tracks your Claude Code and Codex quota usage and pace, so you use every token without running dry.
Categories
Indexed Skills (16)
glass-ui-changes
How to change and verify the look of this app — the desktop card, the menu bar dropdown, the glass panes, borders, rims and the menu bar icon. Use whenever a task touches Glass.swift, PanelView.swift, DesktopCard.swift, UsageCard.swift, MenuBarItem.swift or StatusIcon.swift, or whenever someone says an edge, colour, shadow or border "looks off". Covers the pixel-measurement loop that has to back any claim about why something renders the way it does.
writing-skills
How to write a new skill for this project, or fix an existing one — file layout under .agents/skills, the symlink into .claude/skills, frontmatter rules, what belongs in a skill versus a comment or CLAUDE.md, and how to tell a useful skill from a wasted one. Use when asked to "learn this", "make a skill", "remember how to do X", "write this up so it doesn't happen again", or when a mistake is worth not repeating.
multi-phase-plan
Plan a large change as a checklist of verifiable units before implementing. Use when work spans many files or several PRs and the approach isn't a one-shot. Triggers: "plan this", "break into phases", "multi-PR", "large change", "roadmap", "how should we sequence".
authoring-a-playbook
How to write a new bento playbook (a bento-core skill). Use when adding or editing a playbook, or turning a repeated task into one. Triggers: "add a playbook", "new skill for bento", "capture this as a skill", "write a SKILL.md".
checkpoint
Capture in-progress state so work survives a context reset or handoff. Use when pausing, saving progress, or before a long/risky step. Triggers: "save progress", "checkpoint", "pause here", "resume later", "hand this off".
cso
Security audit of a change or surface — find vulnerabilities before they ship. Use for a security review, or when touching auth, input handling, secrets, or data access. Triggers: "security review", "is this safe", "audit for vulnerabilities", "check for security issues".
docs
Write or update documentation so it matches the current code, and produce release notes. Use after shipping, or when docs have drifted. Triggers: "update the docs", "document this", "write release notes", "changelog", "the docs are out of date".
eval-blind
Compare two models, prompts, or variants without letting bias leak into the verdict. Use when: (1) deciding which model/prompt to ship, (2) grading A vs B outputs, (3) any "is the new version better" comparison. Triggers: "compare models", "which prompt is better", "eval", "A/B", "blind judge".
figure-it-out
Fallback for a task no other playbook covers. Use when the request doesn't match an existing skill and you'd otherwise improvise. Triggers: an off-catalog task, "there's no skill for this", any multi-step work with no matching playbook.
hillclimb
Tune anything measurable — retrieval quality, latency, eval scores, a prompt — by disciplined one-variable-at-a-time experiments with a kept/reverted log. Use when: (1) improving a metric, (2) tuning retrieval/RAG or performance, (3) iterating on a prompt against a dataset. Triggers: "tune", "improve the score", "optimize", "why did it get worse", "iterate on".
investigate
Evidence-first root-cause debugging. Use when something is broken, erroring, or behaving wrong and you need the cause, not a symptom patch. Triggers: "why is this broken", "debug this", "500 error", "it fails when", "find the root cause".
prototype
Build a throwaway to make a design decision cheaply. Use when the right approach is unclear and you need to try before committing. Triggers: "prototype", "spike", "mock it up", "try an approach", "which design/layout", "explore options".
qa
Exercise a change against its real surface and report defects (browser, CLI, or API). Use before shipping UI/flow changes or to find bugs. Triggers: "QA this", "test the site/app", "find bugs", "does this actually work", "smoke test the change".
refactor
Change structure while holding behavior constant. Use when cleaning up, restructuring, or reducing complexity without adding features or fixing bugs. Triggers: "refactor", "clean up", "restructure", "reduce complexity", "tidy this".
ship
Land a change safely: sync base, verify, review the diff, then commit/push/PR. Use when finishing a piece of work. Triggers: "ship it", "land this", "open a PR", "create a PR", "commit and push", "ready to merge".
spec
Turn a vague intent into a precise, executable spec before building. Use when the ask is fuzzy or underspecified and jumping to code would guess wrong. Triggers: "spec this out", "write a spec", "nail down requirements", "what exactly should this do".
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.