📦

videocut

Name: videocut
Author: chengfeng / AI产品自由

Révision du contenu r3 Critique ⚙️ Commandes externes📁 Accès au système de fichiers⚡ Contient des scripts🌐 Accès réseau🔑 Variables d’environnement

Nettoyer des vidéos parlées avec des coupes horodatées

Les vidéos parlées contiennent souvent des erreurs, des mots de remplissage et de longs silences. Cette compétence applique des montages horodatés confirmés avec FFmpeg et vérifie le résultat par transcription.

Prend en charge: Claude Codex Code(CC)

⚠️ 38 Médiocre

Maintain this Skill? Claim it ⚠️ 38 Médiocre

N’installez pas automatiquement ce Skill.

La politique de référence exige un examen par un opérateur avant toute action d’installation.

Installer avec mon Agent

Copiez cette demande dans votre Agent. Elle contient la page Skill de référence et le manifeste.

Demande d’agent

Review the Skillstore skill "videocut" from https://skillstore.io/skills/ceeon-videocut.md and its manifest at https://skillstore.io/api/skills/ceeon-videocut/manifest. Verify the artifact. Do not auto-install. Inspect the skill and report your findings, then wait for an operator or manual installation decision.

Votre Agent doit toujours présenter son plan et demander toute confirmation exigée par la politique de sécurité.

Ressources lisibles par les agents

Utilisez ces liens lorsqu’un Agent IA, un crawler ou un script a besoin d’un contexte propre au lieu de lire toute la page.

Détail Markdown GET /skills/ceeon-videocut.md manifest signé GET /api/skills/ceeon-videocut/manifest lockfile signé GET /api/skills/ceeon-videocut/lockfile

Tester

Utilisation de « videocut ». Une liste de tâches confirmées marque trois erreurs de parole et deux longs silences à supprimer.

Résultat attendu:

La compétence produit une nouvelle version vidéo montée, une transcription de vérification et un fichier de sous-titres SRT.

Utilisation de « videocut ». L'utilisateur demande de conserver deux éléments de silence tout en supprimant toutes les autres plages cochées.

Résultat attendu:

La coupe finale préserve les pauses demandées et supprime uniquement les plages horodatées approuvées.

Utilisation de « videocut ». Une passe de révision trouve un mot de remplissage restant après la première coupe.

Résultat attendu:

Le workflow crée une nouvelle tâche de suppression et répète le montage après confirmation de l'utilisateur.

Audit de sécurité

Critique

v9 • 23/07/2026 Ouvrir le rapport versionné

Most alerts are false positives from Markdown examples, JavaScript template literals, expected media I/O, fixed relative paths, and Chinese text entropy. Confirmed risks include shell injection, an unauthenticated traversal-prone review server, generated-page script injection, arbitrary URL fetching, and external audio and credential handling. Publication should wait until the high-risk execution and server issues are fixed.

Fichiers analysés

3,676

Lignes analysées

Éléments à examiner

Faux positifs ignorés

Préoccupations de sécurité confirmées (4)

Critique

scripts/review_server.js:35-40 scripts/review_server.js:47-139 scripts/review_server.js:142-181 scripts/review_server.js:370

Unauthenticated Review Server Exposes Files and Cut Operations

The server allows any origin, maps request URLs directly to filesystem paths, and exposes unauthenticated write and cut endpoints. Its unspecified listen host can expose traversal-based file reads and expensive edits to network peers.

The code has permissive CORS, no authentication, no path canonicalization, and no loopback host restriction. The exposed handlers perform file writes and invoke media processing.

Élevé

scripts/volcengine_transcribe.sh:45

Environment file access

ENV_FILE="$(dirname "$(dirname "$SCRIPT_DIR")")/.env"

The script resolves and uses a shared .env file outside its own directory. It later reads VOLCENGINE_API_KEY from that credential file.

Élevé

scripts/generate_review.js:778 scripts/generate_review.js:879-880

Generated Review Page Allows Script Injection

The generator inserts the video filename and serialized transcript data directly into HTML and an inline script without HTML-safe escaping. Crafted input can close the script or markup context and execute JavaScript.

The direct template interpolation is visible and JSON.stringify does not escape the HTML script-closing sequence. Both filename and transcript files are caller-controlled inputs.

Élevé

scripts/volcengine_transcribe.sh:72-79 scripts/volcengine_transcribe.sh:108-144

Arbitrary Audio URL Fetch Enables SSRF and Reupload

The transcription script accepts any HTTP or HTTPS URL, follows redirects, and applies no destination restrictions. It then sends the downloaded content to Volcengine, enabling access to internal URLs and external disclosure.

The URL is accepted from the first argument and passed to curl with redirect following. The resulting file is base64-encoded and submitted to a third-party endpoint.

Éléments d’évaluation des capacités (41)

Il s’agit de capacités locales réelles qui peuvent être attendues pour cette compétence ; elles nécessitent donc un examen, mais ne sont pas comptabilisées comme comportement malveillant confirmé.

Élevé

scripts/cut_video.sh:59-60

Shell command substitution

TOTAL_SEGS=$(node -e "

The command substitution builds Node source containing the caller-controlled deletion-list path and ffprobe output. A path containing quotes can escape the JavaScript string and execute code.

Élevé

scripts/cut_video.sh:120

Unix shell invocation

const script = '#!/bin/bash\nffmpeg -y -v error' +

This generated shell script embeds the input video path without shell escaping and is later executed with bash. A crafted filename can break the quoted argument and inject commands.

Élevé

scripts/cut_video.sh:61

Node.js fs operations

const deleteSegs = JSON.parse(fs.readFileSync('$DELETE_JSON', 'utf8'));

The deletion-list path is interpolated directly into Node source rather than passed as an argument. Quote characters in the path create a code-injection path.

Élevé

scripts/cut_video.sh:130

Node.js fs operations

fs.writeFileSync('$TMP_DIR/cmd_' + padded + '.sh', script);

The file operation writes executable shell scripts whose command text contains an unescaped input video path. Those files are executed later, making crafted filenames dangerous.

Élevé

scripts/cut_video.sh:61

Synchronous file operations

const deleteSegs = JSON.parse(fs.readFileSync('$DELETE_JSON', 'utf8'));

The synchronous read is expressed inside dynamically constructed Node source with an unescaped caller-controlled path. This can become JavaScript code execution.

Élevé

scripts/cut_video.sh:130

Synchronous file operations

fs.writeFileSync('$TMP_DIR/cmd_' + padded + '.sh', script);

This write persists generated shell commands containing an unescaped video path. The later bash execution turns malformed filenames into a command-injection risk.

Élevé

scripts/review_server.js:16

Node.js child_process module

const { execSync } = require('child_process');

The server imports child_process and later builds shell command strings from the video path. Because the path can be caller-controlled, the module enables command injection in request-triggered work.

Élevé

scripts/review_server.js:107

Synchronous process execution

execSync(`bash "${scriptPath}" "${VIDEO_FILE}" delete_segments.json "${outputFile}"`, {

execSync runs a shell command string containing the caller-provided video path or an output path derived from it. Quoting alone does not prevent injection through embedded quotes or command substitutions.

Élevé

scripts/review_server.js:114

Synchronous process execution

const originalDuration = parseFloat(execSync(`ffprobe -v error -show_entries format=duration -of csv

Élevé

scripts/review_server.js:115

Synchronous process execution

const newDuration = parseFloat(execSync(`ffprobe -v error -show_entries format=duration -of csv=p=0

Élevé

scripts/review_server.js:242

Synchronous process execution

audioOffset = parseFloat(execSync(`ffprobe -v error -show_entries format=start_time -of csv=p=0 "${a

execSync receives a command string containing a workspace-derived path. A quote or shell metacharacter in that path can alter the command.

Élevé

scripts/review_server.js:251

Synchronous process execution

const duration = parseFloat(execSync(probeCmd).toString().trim());

Élevé

scripts/review_server.js:326

Synchronous process execution

execSync(cmd, { stdio: 'pipe' });

Élevé

scripts/review_server.js:329

Synchronous process execution

const newDuration = parseFloat(execSync(`ffprobe -v error -show_entries format=duration -of csv=p=0

Élevé

scripts/review_server.js:352

Synchronous process execution

execSync(cmd, { stdio: 'pipe' });

Élevé

scripts/review_server.js:362

Synchronous process execution

execSync(concatCmd, { stdio: 'pipe' });

Élevé

scripts/review_server.js:107

Ruby/shell backtick execution

execSync(`bash "${scriptPath}" "${VIDEO_FILE}" delete_segments.json "${outputFile}"`, {

This template literal is used to construct a shell command with a caller-controlled or derived media path. The resulting string is executed by a shell without robust escaping.

Élevé

scripts/review_server.js:114

Ruby/shell backtick execution

const originalDuration = parseFloat(execSync(`ffprobe -v error -show_entries format=duration -of csv

This template literal is used to construct a shell command with a caller-controlled or derived media path. The resulting string is executed by a shell without robust escaping.

Élevé

scripts/review_server.js:115

Ruby/shell backtick execution

const newDuration = parseFloat(execSync(`ffprobe -v error -show_entries format=duration -of csv=p=0

This template literal is used to construct a shell command with a caller-controlled or derived media path. The resulting string is executed by a shell without robust escaping.

Élevé

scripts/review_server.js:242

Ruby/shell backtick execution

audioOffset = parseFloat(execSync(`ffprobe -v error -show_entries format=start_time -of csv=p=0 "${a

This JavaScript template literal constructs a shell command with a workspace-derived path and then passes it to execSync. A specially named directory can inject shell syntax.

Élevé

scripts/review_server.js:250

Ruby/shell backtick execution

const probeCmd = `ffprobe -v error -show_entries format=duration -of csv=p=0 "file:${input}"`;

This template literal is used to construct a shell command with a caller-controlled or derived media path. The resulting string is executed by a shell without robust escaping.

Élevé

scripts/review_server.js:323

Ruby/shell backtick execution

const cmd = `ffmpeg -y -i "file:${input}" -filter_complex "${filterComplex}" -map "[outv]" -map "[ou

This template literal is used to construct a shell command with a caller-controlled or derived media path. The resulting string is executed by a shell without robust escaping.

Élevé

scripts/review_server.js:329

Ruby/shell backtick execution

const newDuration = parseFloat(execSync(`ffprobe -v error -show_entries format=duration -of csv=p=0

This template literal is used to construct a shell command with a caller-controlled or derived media path. The resulting string is executed by a shell without robust escaping.

Élevé

scripts/review_server.js:349

Ruby/shell backtick execution

const cmd = `ffmpeg -y -ss ${seg.start.toFixed(3)} -i "file:${input}" -t ${segDuration.toFixed(3)} -

This template literal is used to construct a shell command with a caller-controlled or derived media path. The resulting string is executed by a shell without robust escaping.

Élevé

scripts/review_server.js:360

Ruby/shell backtick execution

const concatCmd = `ffmpeg -y -f concat -safe 0 -i "${listFile}" -c copy "${output}"`;

This template literal is used to construct a shell command with a caller-controlled or derived media path. The resulting string is executed by a shell without robust escaping.

Élevé

scripts/review_server.js:53

Node.js fs operations

fs.writeFileSync('saved_selection.json', body);

An unauthenticated HTTP endpoint accepts an unbounded request body and writes it to a fixed workspace file. Network clients can overwrite review state and consume memory or disk.

Élevé

scripts/review_server.js:91

Node.js fs operations

fs.writeFileSync('delete_segments.json', JSON.stringify(deleteList, null, 2));

The unauthenticated cut endpoint writes attacker-supplied deletion ranges and immediately starts media processing. Any reachable client can alter output and trigger expensive work.

Élevé

scripts/review_server.js:150

Node.js fs stat operations

if (!fs.existsSync(filePath)) {

The server applies the request URL directly to a relative filesystem path without canonicalization or root containment. Traversal sequences can make the stat operation target files outside the review directory.

Élevé

scripts/review_server.js:156

Node.js fs stat operations

const stat = fs.statSync(filePath);

Élevé

scripts/review_server.js:53

Synchronous file operations

fs.writeFileSync('saved_selection.json', body);

An unauthenticated HTTP endpoint accepts an unbounded request body and writes it to a fixed workspace file. Network clients can overwrite review state and consume memory or disk.

Élevé

scripts/review_server.js:91

Synchronous file operations

fs.writeFileSync('delete_segments.json', JSON.stringify(deleteList, null, 2));

The unauthenticated cut endpoint writes attacker-supplied deletion ranges and immediately starts media processing. Any reachable client can alter output and trigger expensive work.

Élevé

scripts/review_server.js:171

File stream creation

fs.createReadStream(filePath, { start, end }).pipe(res);

The stream serves a path derived directly from req.url without enforcing the review directory as a root. A traversal request can disclose arbitrary readable local files.

Élevé

scripts/review_server.js:181

File stream creation

fs.createReadStream(filePath).pipe(res);

The stream serves a path derived directly from req.url without enforcing the review directory as a root. A traversal request can disclose arbitrary readable local files.

Élevé

scripts/volcengine_transcribe.sh:18

Hardcoded URL

SUBMIT_URL="https://openspeech.bytedance.com/api/v3/auc/bigmodel/submit"

This endpoint receives the user audio and API key over HTTPS as part of transcription. The transfer is intentional but is a real external data-disclosure boundary.

Élevé

scripts/volcengine_transcribe.sh:19

Hardcoded URL

QUERY_URL="https://openspeech.bytedance.com/api/v3/auc/bigmodel/query"

This endpoint is polled with the API key to retrieve transcription results. It is an intentional third-party network dependency with credential exposure.

Élevé

scripts/volcengine_transcribe.sh:111

Node.js fs operations

data: fs.readFileSync(audioFile).toString('base64'),

The script reads the complete audio file and base64-encodes it into a request sent to Volcengine. This is intentional, but user media leaves the local environment.

Élevé

scripts/volcengine_transcribe.sh:111

Synchronous file operations

data: fs.readFileSync(audioFile).toString('base64'),

The script reads the complete audio file and base64-encodes it into a request sent to Volcengine. This is intentional, but user media leaves the local environment.

Élevé

scripts/volcengine_transcribe.sh:45

Hidden file access

ENV_FILE="$(dirname "$(dirname "$SCRIPT_DIR")")/.env"

The script locates a shared ancestor .env file and later extracts a live API key from it. This is real access to hidden credential storage.

Élevé

scripts/volcengine_transcribe.sh:54

Generic API/secret keys

API_KEY=$(grep '^VOLCENGINE_API_KEY=' "$ENV_FILE" | tail -1 | cut -d'=' -f2- | tr -d '"' | tr -d "'"

This line reads VOLCENGINE_API_KEY from the shared .env file into process memory. The access is required for transcription but handles a live secret.

Élevé

scripts/volcengine_transcribe.sh:140

Generic API/secret keys

-H "X-Api-Key: $API_KEY" \

The API key is placed in an outbound X-Api-Key header to the configured Volcengine endpoint. This is expected authentication but remains credential disclosure to a third party.

Élevé

scripts/volcengine_transcribe.sh:166

Generic API/secret keys

-H "X-Api-Key: $API_KEY" \

The API key is placed in an outbound X-Api-Key header to the configured Volcengine endpoint. This is expected authentication but remains credential disclosure to a third party.

Facteurs de risque

⚙️ Commandes externes (50)

📁 Accès au système de fichiers (50)

⚡ Contient des scripts (5)

scripts/generate_review.js:895 scripts/generate_review.js:896 scripts/generate_review.js:1017 scripts/generate_review.js:1027 scripts/generate_review.js:1158

🌐 Accès réseau (12)

scripts/generate_review.js:1279 scripts/generate_review.js:1289 scripts/generate_review.js:1427 scripts/generate_review.js:1484 scripts/review_server.js:373 scripts/volcengine_transcribe.sh:7 scripts/volcengine_transcribe.sh:17 scripts/volcengine_transcribe.sh:18 scripts/volcengine_transcribe.sh:19 SKILL.md:5 SKILL.md:362 SKILL.md:515

🔑 Variables d’environnement (10)

scripts/volcengine_transcribe.sh:23 scripts/volcengine_transcribe.sh:29 scripts/volcengine_transcribe.sh:49 scripts/volcengine_transcribe.sh:54 scripts/volcengine_transcribe.sh:56 scripts/volcengine_transcribe.sh:57 scripts/volcengine_transcribe.sh:58 scripts/volcengine_transcribe.sh:140 scripts/volcengine_transcribe.sh:166 SKILL.md:530

Audité par: codex Voir l’historique des audits →

Partager et citer ce rapport

Partagez le rapport d’évaluation versionné, le badge neutre, la carte intégrable et les citations. Skillstore présente les preuves sans décider si ce Skill est sûr.

Ouvrir le rapport versionné

Copier le lien du rapport

https://skillstore.io/skills/ceeon-videocut/audits/9?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Badge Markdown

[![Skillstore security assessment](https://skillstore.io/badges/skills/ceeon-videocut/security.svg)](https://skillstore.io/skills/ceeon-videocut?utm_source=security_passport_badge)

Badge HTML

<a href="https://skillstore.io/skills/ceeon-videocut?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/ceeon-videocut/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Carte intégrable

<iframe src="https://skillstore.io/embed/skills/ceeon-videocut.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

Citations académiques (APA · BibTeX · CFF)

Citation APA

chengfeng / AI产品自由. (2026). videocut security audit report (audit version 9) [Author version unspecified]. Skillstore. https://skillstore.io/skills/ceeon-videocut/audits/9

Citation BibTeX

@techreport{chengfeng-ai-ceeon-videocut-2026,
  author = {chengfeng / AI产品自由},
  title = {videocut security audit report (audit version 9)},
  institution = {Skillstore},
  year = {2026},
  number = {9},
  url = {https://skillstore.io/skills/ceeon-videocut/audits/9},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "videocut security audit report (audit version 9)"
version: "unspecified"
type: report
authors:
  - name: "chengfeng / AI产品自由"
date-released: "2026-07-23"
url: "https://skillstore.io/skills/ceeon-videocut/audits/9"
identifiers:
  - type: other
    value: "skillstore:ceeon-videocut:audit:9"
    description: "Skillstore immutable audit report identifier"

Score Skillstore

Pourquoi ce score Niveau de confiance des preuves: Moyen

Architecture

Maintenabilité

Contenu

Communauté

Conformité aux spécifications

Ce que vous pouvez créer

Monter des vidéos de créateurs face caméra

Supprimer les erreurs confirmées, les mots de remplissage et les longues pauses des vidéos parlées enregistrées.

Préparer des extraits de cours

Appliquer des coupes horodatées précises aux enregistrements de cours avant de publier du matériel pédagogique.

Nettoyer des briefings internes

Transformer des enregistrements bruts de réunions ou de briefings en extraits plus propres avec sous-titres.

Essayez ces prompts

Exécuter les coupes confirmées

Utilise videocut pour exécuter les tâches de suppression confirmées pour cette vidéo. Conserve les éléments non cochés et génère la vidéo montée.

Couper et vérifier la parole

Coupe les plages horodatées cochées, retranscris le fichier monté et indique s'il reste des erreurs de parole.

Préserver certains silences

Exécute toutes les coupes cochées sauf les éléments de silence 3 et 5. Conserve ces pauses, puis crée des sous-titres pour la version finale.

Itérer jusqu'à obtenir une version propre

Applique la liste de suppression actuelle, crée une nouvelle version, relance la révision de transcription et poursuis uniquement avec les nouvelles coupes confirmées.

Bonnes pratiques

Confirmer chaque plage de suppression avant d'exécuter FFmpeg.
Utiliser des plages horodatées au lieu de rechercher le texte de la transcription.
Relire la transcription montée avant de publier la vidéo finale.

Éviter

Ne supprimez pas les tâches non cochées de la liste de révision.
Ne relancez pas une recherche de texte lorsque des horodatages précis existent déjà.
N'exécutez pas de commandes multimédias sur des chemins non fiables sans confirmation de l'utilisateur.