Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add kyh/vibedgames --skill finish-itgit clone --depth 1 https://github.com/kyh/vibedgamesWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/kyh/vibedgames/finish-it)<a href="https://agentmods.dev/skills/kyh/vibedgames/finish-it"><img src="https://agentmods.dev/badge/skills/kyh/vibedgames/finish-it/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/kyh/vibedgames/finish-it"><img src="https://agentmods.dev/badge/skills/kyh/vibedgames/finish-it.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00029 | $0.01345 |
| Opus 5.5 | $0.00012 | $0.00538 |
| Sonnet 5.5 | $0.00006 | $0.00269 |
| Haiku 4.5 | $0.00003 | $0.00135 |
Grade A, and why
finish-it scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 101 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Finish it
Finishing is a separate skill, and a finished mediocre game proves more than a brilliant fragment. Projects die in two loops: the restart loop (skills improved mid-project, so the dev remakes early content sideways forever — "your code will always be a mess") and the polish loop (near-done, endlessly refining details out of fear of judgment). Diagnose the loop before intervening. And remember: the last 10% is really 90% — menus, death states, audio plumbing, deploy are the game too.
Scoping a new game (so this skill is never needed)
- One core mechanic, one level/arena, one win condition, one lose condition. Everything else is stretch.
- One-sentence pitch before code. Features not in the sentence need explicit justification.
- Budget 50/50: half for the playable loop, half for the unfun tail. If the loop fills the budget, the scope is wrong.
- Prototype-first split (jam culture): playable ugly loop in the first ~5% of the time; the rest is feel, art, sound, content.
- Use the engine at hand (Spelunky shipped in Game Maker), reuse and generate assets without shame.
- Tiny finished games grow up: Celeste began as a 4-day PICO-8 jam game. The jam-sized version is the validation step, not a compromise.
The cut list
- Keep a
NEXT_GAME.md. Every mid-build idea ("what if it had shops / multiplayer / procgen") goes there, not in the build. Save it for the next one — finish this one first (Yu). - Default cut order for browser games: extra levels → extra enemy/item types → meta-progression → settings menus → narrative → leaderboards.
- Never cut: restart flow, lose/win states, audio toggle.
- When behind: cut features. One mechanic deep beats three shallow.
- If a project must die, the successor is smaller, not bigger.
When is the core loop enough?
Stop adding mechanics when: a new player understands it without instructions
in <30s, a session has tension (can lose) and release (can win/score), and a
player voluntarily restarts once. From that moment, every proposal to add is
scope creep — route it to NEXT_GAME.md and switch to finishing mode.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · +3 lines · -86 tokens per session 2ce89f757834
- 1mo ago First seen · 98 lines · 115 tokens per session scan A e743f4c7634e
finish-it is a skill published in the GitHub repository kyh/vibedgames (61 stars, last pushed yesterday), licensed MIT. It adds 29 tokens to every session and 1,345 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
inspector-workbench
Runs Inspector UI work on the standalone Threads state lab so the agent can see the overlay. Use when a CopilotKit employee asks an agent to fix, polish, add, or debug Inspector UI, chrome, panes, overlay actions, launcher, Home, Threads, Playground, Memory, or any visual behavior in @copilotkit/web-inspector. Don't…
dev-fix
Unified developer workflow for fixing bugs. Analyzes issue-tracker context, cross-checks docs/code, proposes a solution, implements the fix, verifies locally, and delivers a PR/MR.
verify-bug
Post-merge UAT verification workflow. Walks JIRA reproduce steps, performs comparative audits (Before/After), attaches evidence to JIRA, and transitions status on PASS.
incident-hotfix
Mitigate a production incident or urgent regression first, then route to root-cause remediation and a postmortem.
cyber-triage
Evidence-led blue incident triage with authorization and runtime gates.
common-performance-engineering
Enforce universal standards for high-performance development. Use when profiling bottlenecks, reducing latency, fixing memory leaks, improving throughput, or optimizing algorithm complexity in any language.