Agent Economy Report

Skill code audit

SAFE version 1.4.0 · audited 2026-09-09

remotion-video-toolkit @shreefentsar

This package is a documentation-only skill: a set of Markdown guides plus two example React/TSX files that teach an AI agent how to write Remotion (React-based) video code and how to render videos. It contains no scripts that run on install, no network calls made by the skill itself, and no access to credentials, wallets or payment systems. The shell commands it shows are standard, publicly documented Remotion/npm commands (`npx create-video`, `npx remotion render`, `npx remotion add @remotion/...`) that an agent would only run if the user asks it to scaffold or render a video. The external domains referenced are documentation links (remotion.dev, github.com, clawhub.ai, publisher site 99.zone) or obvious placeholders in code samples (example.com, api.example.com, remotion.media, lottiefiles). Nothing in the files tries to override an agent's rules, exfiltrate data, or persist itself. The main things to be aware of are ordinary, disclosed risks: the guides suggest installing npm packages and deploying cloud rendering (AWS Lambda / Google Cloud Run), and one guide tells the agent to fetch a Remotion docs page from the web.

Declared purposeA toolkit of 29 rule documents teaching an AI agent how to build programmatic videos with Remotion + React: animations, timing, rendering (CLI/Node/Lambda/Cloud Run), captions, charts, 3D, text effects, transitions and media handling.
Observed behaviorThe package is entirely reference material. Every file is Markdown guidance or an illustrative .tsx component that computes animation values from Remotion hooks. No file executes anything on its own: there are no install hooks, no exec/spawn calls, no writes to disk, no reads of environment secrets, and no outbound requests initiated by the skill. Shell snippets are inert text inside code fences and match Remotion's official CLI. The only sensitive-adjacent guidance is standard: run npm/npx package installs, deploy render workers to AWS Lambda or Google Cloud Run, and fetch a remotion.dev docs page for Tailwind setup.
Verdict, rules onlySAFE
Verdict, AISAFE
Final verdictSAFE (never better than either pass)

Capabilities

capabilityvs purposedetail
shellneededDocumentation instructs the agent to run standard Remotion CLI commands, e.g. SKILL.md line 129 `npx create-video@latest my-video` and rules/rendering.md line 15 `npx remotion render ...`. These are text instructions, not executed code.
install_packagesneededMany rules tell the agent to install official Remotion packages via npm/bun/yarn/pnpm, e.g. rules/videos.md line 16 `npx remotion add @remotion/media`. All package names are first-party @remotion scopes; no unknown or typosquat packages.
networkneededGuidance references remote media URLs and fetching data at render time (rules/compositions.md line 117 fetch to api.example.com placeholder; rules/lottie.md line 42 fetch from assets4.lottiefiles.com). rules/tailwind.md line 11 asks the agent to WebFetch the official remotion.dev docs page. No hidden or attacker-controlled endpoints.
filesystem_writeneededRendering guidance produces output files in the user's project (e.g. `out/video.mp4`); the skill itself contains no file-write code.
credentialsexceeds purposeNo credential reads. The only env var found is a cosmetic color constant COLOR_AXIS in rules/assets/charts-bar-chart.tsx line 10. Cloud rendering guidance implies the user's own AWS/GCP credentials are used by the Remotion CLI, but the skill never reads or transmits them.

Findings

severityfinding
lowInstructs the agent to fetch and follow instructions from an external web page
The skill delegates setup instructions to a live third-party page (the official Remotion docs). This is a reasonable source, but any content pulled from the web becomes untrusted input in the agent's context, and the instructions there could change over time without the skill being updated.
rules/tailwind.md:11 Tailwind must be installed and enabled first in a Remotion project - fetch https://www.remotion.dev/docs/tailwind using WebFetch for instructions.
lowGuidance runs npx against a remote package at latest version
Following the quick start causes the agent to download and execute the current release of the create-video scaffolder from the npm registry. This is the standard, official Remotion setup path, but it is unpinned download-and-run behavior that depends on npm supply-chain integrity.
SKILL.md:129 npx create-video@latest my-video
infoPackage installation instructions
Multiple rule files instruct installation of first-party @remotion/* packages (media, captions, three, gif, fonts, google-fonts, transitions, layout-utils, lottie). All are official Remotion scoped packages; no unexpected or lookalike dependencies were found.
rules/videos.md:16 npx remotion add @remotion/media # If project uses npm
lowCloud deployment guidance touches the user's AWS/GCP accounts
The rendering guide walks the agent through creating IAM policies and deploying Lambda functions and Cloud Run services. Executed with real cloud credentials this creates billable infrastructure and modifies account permissions. The skill does not read or transmit credentials itself, but a user should expect real spend and IAM changes if the agent follows these steps.
rules/rendering.md:137 npx remotion lambda policies role
mediumExample HTTP render endpoint has no authentication or input validation
The sample Express server takes a caller-supplied compositionId and arbitrary props and renders them, with no auth, rate limiting, or schema validation. If an agent copies this pattern into production it exposes an unauthenticated, CPU-intensive endpoint. The package's own skill-card.md lines 19-22 acknowledges this risk and recommends mitigations.
rules/rendering.md:210 app.post("/render", async (req, res) => {
infoNetwork fetch in example uses an obvious placeholder host
The pre-render data fetch example points at api.example.com, a reserved documentation domain, not a live endpoint. No hidden or hard-coded exfiltration destination is present anywhere in the package.
rules/compositions.md:117 const data = await fetch(`https://api.example.com/video/${props.videoId}`, {

0 AI finding(s) were dropped because their file, line or quote did not match the package.

Network destinations in the code

domainservice ratingwhere
99.zonenot in the service indexSKILL.md:149
api.example.comnot in the service indexrules/compositions.md:117
assets4.lottiefiles.comnot in the service indexrules/lottie.md:42
clawhub.ainot in the service indexskill-card.md:7
example.comnot in the service indexrules/assets.md:71
github.comnot in the service indexSKILL.md:145
remotion.devnot in the service indexrules/transcribe-captions.md:13
remotion.medianot in the service indexrules/assets.md:72
storage.googleapis.comnot in the service indexrules/rendering.md:183
www.remotion.devnot in the service indexrules/tailwind.md:11

Other facts

Binaries invokedbash, npm, npx
Environment variables readCOLOR_AXIS
Hard-coded walletsnone
Pipes a download to a shell0
eval / exec / subprocess0
base64 blobs0
File writes0
Persistence0

Files audited

filelinessha256
SKILL.md149140a3e508a447954…
_meta.json60d3cf66ee7dd8663…
rules/3d.md862a5b1a72de35168b…
rules/animations.md2999a6377d64061af5…
rules/assets.md78664160a48fe6ef76…
rules/assets/charts-bar-chart.tsx1736e5e219c596c8515…
rules/assets/text-animations-typewriter.tsx10092b5f47d6d18387a…
rules/assets/text-animations-word-highlight.tsx1085dbba688b875baaa…
rules/audio.md17206d859b2f463d0e0…
rules/calculate-metadata.md104b258c788fb333718…
rules/can-decode.md75fb3f9933e45103d7…
rules/charts.md582430705a627c468e…
rules/compositions.md14658d92350fecc1cb5…
rules/display-captions.md12696abde72e0bcb88c…
rules/extract-frames.md229b7a63f6509a1a33d…
rules/fonts.md1521749e4cf82e8b157…
rules/get-audio-duration.md58d5098e8d20b22746…
rules/get-video-dimensions.md68839867aa115e09a7…
rules/get-video-duration.md587b31fab3cbf9b077…
rules/gifs.md13853d9d658e3093dba…
rules/images.md13091d32f6e528b1f1c…
rules/import-srt-captions.md6738d33be5439714a2…
rules/lottie.md686c0cb4babcff94e3…
rules/measuring-dom-nodes.md355270db465cb9eed1…
rules/measuring-text.md14388905cbd701e2473…
rules/rendering.md2414630b82b206e0976…
rules/sequencing.md106ecb74589c87da53d…
rules/tailwind.md11bdd5162400cd20c6…
rules/text-animations.md2040ac6b1bccf1c57d…
rules/timing.md179f639ed8452767a74…
rules/transcribe-captions.md19e7d3edf667e3b1f3…
rules/transitions.md122a42100803a1e5e67…
rules/trimming.md5340727aefd5939a42…
rules/videos.md171d9d3032a57f92c75…
skill-card.md42fc6f1fd36ce8db16…

For agents

JSON: https://agenteconomy.report/k/remotion-video-toolkit.audit.json · badge: https://agenteconomy.report/k/remotion-video-toolkit.audited.svg · skill rating: /k/remotion-video-toolkit · commission an audit of any skill: US$ 29 per version.

How this is computed

The complete published package of this exact version was downloaded from the registry and read statically; nothing was executed. A deterministic pass extracts network destinations, binaries, environment variables, writes, install commands, obfuscation markers, subprocesses and wallets, each with file and line. An AI then reads the whole package with those facts and writes the summary, the capabilities and the findings under a strict schema; every finding it produces must cite a file, a line and the exact text, or it is dropped. The final verdict is the worst of the two passes. The audit does not change the skill's trust tier (policy); the author may respond through the dispute channel and the response is published here. Commissioned by: the Agent Economy Report (free program: the 150 most downloaded skills, September 2026).