simantaturja
UserLocal, zero-setup dashboard for Claude Code token usage and cost — per project, client, model, and prompt, with an efficiency advisor.
Categories
Indexed Skills (8)
issue-to-pr
Triage the GitHub issues on this repo (cccost-dashboard), judge which are real bugs or worth-building features, then implement the one you pick — TDD, green CI, branch, PR. Use when the user says "check the issues", "triage the issues", "what's in the backlog", "anything worth fixing", "implement issue 12", "fix #12", "work on that issue", or pastes an issue URL. Also use when the user asks whether a reported issue is real, reproducible, or in scope.
release
Cut a new cccost-dashboard release — version bump, CHANGELOG entry, release commit, git tag, GitHub release, npm publish (CI), and VS Code Marketplace publish. Use when the user says "new release", "cut a release", "ship vX.Y.Z", "bump the version", "publish to npm", "publish the extension", or asks what version the next release should be.
cccost-product-owner
Act as the product owner for this repo (cccost-dashboard, the local Claude Code cost dashboard) — a senior PM who knows the LLM-agent cost/usage-tracking category cold. Use whenever the work is about WHAT to build or WHY, not how: proposing or evaluating features, deciding what to build next or cut, sizing an idea against the roadmap, tearing down a competitor (ccusage, Claude Code's /cost & /usage, usage monitors, LiteLLM/Bifrost/Helicone/Mavvrik, Cursor/Copilot dashboards), finding gaps in what other tools do, writing a PRD or feature spec, positioning/differentiation, or prioritization. Trigger on phrasings like "what should we build", "is X worth building", "how do we compare to <tool>", "what are we missing", "find the gaps", "feature idea", "roadmap", "prioritize this", or a one-line feature request with no problem statement. Reach for it even when the user names no artifact but is clearly reasoning about product direction for this dashboard.
loop-design
Phase 5 of the loop. Authors a spec and an implementation plan for a feature issue, commits both to a loop branch, and opens a docs-only draft PR that stops at gate 2. Use when `loop-issue` classifies an issue as a feature, when the user says "design issue 12", "write the spec and plan for this", or when a maintainer comments on a design PR asking for revisions.
loop-issue
Lane 2 of the loop. Takes exactly one GitHub issue that a maintainer has labelled `loop:go`, classifies it, and either builds it (bug or docs) or hands it to `loop-design` (feature). Use when the user says "run the autopilot on #12", "work the labelled issue", "loop this issue", or when the label-gated workflow fires. Acts on one issue only — it does not sweep the backlog.
loop-pipeline
The shared mechanics every loop lane repeats: dedup a finding, open an isolated worktree, hand the work to loop-drafter, run npm run verify, hand the diff to loop-verifier, open a draft PR, and append to .loop/log.md. Use when a loop lane skill (loop-watchdog, loop-issue) has decided a finding is real and needs it carried to a draft PR. Not a triage skill — it does not decide what is worth doing, only what happens next.
loop-plan
Phase 6 of the loop. Executes an approved implementation plan task-by-task after a maintainer applies `loop:build` to its design PR. Delegates to superpowers:subagent-driven-development, runs `npm run verify` between tasks, and ticks the plan's checkboxes as it goes. Use when a design PR is labelled `loop:build`, or when the user says "execute the approved plan", "build the design in PR 12", or "run the plan".
loop-watchdog
Lane 1 of the loop. Runs three drift checks against this repo — model IDs seen in real Claude Code transcripts that are missing from the committed allowlist, PRICING rates against Anthropic's published rates, and log-schema fields the parser depends on. Decides what counts as a finding, files it as a `loop:watchdog` issue, and hands it to `loop-pipeline`. Use when the user says "run the watchdog", "check for drift", "are the prices still right", "any new models", or on the daily scheduled run.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.