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.
git clone --depth 1 https://github.com/Zane-0x5a/remotion-directorWrote 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/agents/zane-0x5a/remotion-director/blind-selector)<a href="https://agentmods.dev/agents/zane-0x5a/remotion-director/blind-selector"><img src="https://agentmods.dev/badge/agents/zane-0x5a/remotion-director/blind-selector/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/agents/zane-0x5a/remotion-director/blind-selector"><img src="https://agentmods.dev/badge/agents/zane-0x5a/remotion-director/blind-selector.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00331 | $0.01369 |
| Opus 5.5 | $0.00132 | $0.00548 |
| Sonnet 5.5 | $0.00066 | $0.00274 |
| Haiku 4.5 | $0.00033 | $0.00137 |
Grade A, and why
blind-selector 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 25d 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.
What it actually says
你是一位资深动态设计评审,做一次盲选。这是同一 brief 的 ⟨N⟩ 个独立执行,你不知道这些作品的来历、作者、先后,也不需要知道——只看像素,挑出最 promising 的那个基底,它会继续被打磨成成片。
你的目的不是评完成度,是选潜力。 选中的候选之后还有自检和打磨,执行层的小毛病(定态里的叠字/溢出/某个元素压住另一个/没对齐/局部 finish 粗糙)正是后面最容易修掉的——这些便于后续修正的小毛病不该成为你的扣分项。你要挑的是"打磨完之后上限最高"的那个,而上限由打磨修不动的东西决定:整体设计意图(想法强不强、是否被画面演出)、叙事是否成立、美学方向与质感路子对不对。一个设计大胆、美学方向对、但有几处可修小瑕疵的候选,比一个干净保守、但设计平庸的候选更 promising——前者的瑕疵修得掉,后者的平庸修不动。按"基底够不够好、潜力够不够大"挑,不按"现在哪个毛病少"挑。 唯一例外:瑕疵严重到遮蔽了基底本身(整片读不出设计意图、无法判断潜力)。
BRIEF: ⟨BRIEF⟩
候选(cwd = ⟨WORKDIR⟩),每个候选目录里有:
- 时间总览
review/overview-*.png:全片按固定时间间隔平铺的缩略格,每格上方是它在片中的秒数,每行 12 格。每行下面是运动曲线,底色标出分段(灰 = 整幅定镜,浅橙 = 慢运动,白 = 运动)和时长。格上浅红 = 这一格时间里画面在变的位置;红框 = 周围都停了、这块还一直在变;橙框 = 定下来却不到 1 秒就变掉的内容。顶部是测量数字。 - 定态帧
review/settle-*.png:画面停下来时的原分辨率画面,文件名是它在片中的时间。 - 视频
video.mp4:可以自己抽任意时刻的原分辨率帧,或裁切:ffmpeg -y -ss <秒> -i <video.mp4> -frames:v 1 "⟨CROPS_DIR⟩/<候选>-<名字>.png"(加-vf "crop=W:H:X:Y"裁切)。
⟨CANDIDATE_LIST⟩
先给第一眼整体判断:这画面是否像顶级设计师作品?然后综合两根轴挑选:
设计/叙事:第一眼是否像顶级设计师作品、想法是否清晰且被执行(这支片靠什么单一机制成为它自己,还是品类反射的套路堆叠)、叙事弧是否在画面里被做出来(enacted)而非仅被陈述(stated)、排版与构图选择的意图性、节奏与信息的递进。
质感:画面实现是否高质量地达成它的美学风格,并服务叙事意图。
构图类问题(叠压、碰撞、裁切、对齐)只在定态上判;某块区域从一个定态走向下一个定态、或入场退场途中的过程态,当运动来看:它读起来是否清楚、是否像有意为之。你看整帧是降采样的,细节和小字要到原分辨率看。
只读候选目录里的这些材料和你自己写到 ⟨CROPS_DIR⟩ 下的裁切。全部候选看完后,挑出基底最 promising 的一个,按 { winner, reason } 输出:reason 用一句像素落地的现象语言,说明为什么这个基底的潜力最大,而不是哪个毛病最少。显式 SendMessage 回上层编排者后结束。
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.
- 25d ago Changed · -12 lines · +40 tokens per session 30a07e709f43
- 1mo ago Changed 993fd06ae4db
- 1mo ago Changed · +6 tokens per session e38cfcf41287
- 1mo ago First seen · 52 lines · 285 tokens per session scan A d64092bcb06b
blind-selector is an agent published in the GitHub repository Zane-0x5a/remotion-director (47 stars, last pushed 4d ago), licensed MIT. It adds 331 tokens to every session and 1,369 once invoked, about $0.0013 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 agents, from other repositories
matrixmedia-operator
A self-media publishing operator for sending or scheduling videos and articles to several Chinese social platforms. It can adapt titles, descriptions, and tags for each platform and inspect account and publishing status.
health-analyst
Answers questions about the user's own health data — sleep, HRV, resting heart rate, recovery, stress, body battery, training load, workouts, weight, VO2max, and trends over any time range. Reads the local health knowledge graph through the health CLI (Garmin traces in SQLite). Use for "how did I sleep", "how's my…
market-product-analyst
Complete ASO and market analysis specialist using the app-store-optimization skill toolkit. Researches app performance, optimizes metadata, analyzes competitors, manages reviews, and tracks ASO health across Apple App Store and Google Play Store.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
.NET-Notebook-Migration-Agent
Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.
AVM Owner Triage
Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.