Roblox Hit Evaluator: Evidence-Based Breakout Scoring

Where this fits

You ask an agent: 'Should we invest two weeks in this Roblox game before it breaks out?' A generic agent pattern-matches to 'it has lots of visits' and returns a confident probability. The roblox-hit-evaluator skill forces a different path: real collection, observable facts, no fake history, and an auditable score with visible missing-data bounds. It fits any agent making go/no-go judgments from volatile public platform data.

Why agents benefit

• Every score is auditable: each observed number carries source, collection time, and observation window; qualitative tiers link to specific evidence.
• Missing data is not hidden: unknown inputs stay null, not zero, and score bounds remain visible instead of being reweighted away.
• Facts and forecasts are separated: current API responses are never used to fake past-week curves, and view counts never become first-24-hour stats.
• It does real collection, not a plan: the skill's reference files require the agent to gather public evidence before answering.
• Outcome reviews close the loop: references/outcomes.md lets the agent later check whether the previous judgment held.

Practical scenarios

Pre-production greenlight for a small studio

An agent evaluates a social horror game for repeat-play mechanics. It scores repeat-play motivation, differentiation, and spread from public evidence, then lists the retention proxies it could not verify.

Avoiding a content-farm detour

An analyst asks whether an AI-generated obstacle course is worth covering. The skill keeps SEO search volume and domain availability out of the breakout score, so the agent returns a low score with nulls instead of padding with unrelated metrics.

Tracking a previous hot take

A team bet on a roleplay game three weeks ago. The agent reruns the skill in outcome-review mode, compares the frozen forecast with current data, and reports which signals held and which were over-weighted.

Add it to your agent workflow

Install with:

npx skills add kennyzir/7deer_skills --skill roblox-hit-evaluator

Then ask the agent to run it on a specific game:

Run roblox-hit-evaluator on https://www.roblox.com/games/123456/Example
Return the 30-day breakout score, evidence table, and missing-data bounds.

The skill resolves reference files from its own directory, so it works even if the current working directory is different.

Compare before adopting

• Quality signals: check whether scoring rules define null handling and source requirements. This skill does; many ad-hoc prompts do not.
• Maintenance freshness: watch the repo and version. The qilu-hit-v0.1 version suggests iteration, but confirm activity.
• Alternatives: generic web search, a dashboard export, or a hand-written rubric. Compare whether they enforce missing-data bounds and outcome reviews.
• Workflow fit: this is a security/evidence-integrity skill for Roblox breakout judgment, not SEO or domain investing. Use it for short-horizon game opportunity calls.

Why it is worth tracking

This skill models an important discipline: it rejects fake precision, separates observation from projection, and refuses to convert unknown data into zeros. That makes it valuable beyond Roblox as a pattern for auditable agent research. With 309 stars and a versioned workflow, it is worth tracking if you build agents that give go/no-go advice from volatile public data. For an always-updated view and related skills, check https://www.openagentskill.com/skills/kennyzir-roblox-hit-evaluator.