Back to skill

Security audit

丢首歌自动剪卡点视频

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious with medium confidence, chiefly because the scanners report broad advertising and narration features beyond the stated beat-synchronized video purpose, along with privacy and installation concerns.

The verdict is suspicious, so consider avoiding installation until the package’s scope and data handling are clarified. If you proceed, review the need for its cloud services and local-project access, avoid sending sensitive footage or text to external providers, and verify dependency versions and any antivirus alerts before installing.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (200)

Tainted flow: 'headers' from os.environ.get (line 127, credential/environment) → requests.post (network output)

Critical
Category
Data Flow
Confidence
90% confidence
Finding

Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Content

Scanner excerpt · engine/adflow/engine/minimax_tts.py (reported line 135)May include surrounding context.

python
logger.info("MiniMax TTS: %s voice=%s", text[:40], self.voice)

        try:
            r = requests.post(self.API_URL, json=body, headers=headers, timeout=30)
            data = r.json()
        except Exception as e:
            raise RuntimeError(f"MiniMax API 请求失败: {e}")

Tainted flow: 'request_body' from os.environ.get (line 144, credential/environment) → requests.post (network output)

Critical
Category
Data Flow
Confidence
90% confidence
Finding

Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Content

Scanner excerpt · engine/adflow/engine/volcano_tts.py (reported line 177)May include surrounding context.

python
text[:40], self.voice, emotion, speed_ratio)

        try:
            response = requests.post(
                self.API_URL,
                json=request_body,
                headers={

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code has a materially different primary purpose: reference-driven advertising analysis, not music-driven montage creation. Although it inspects video pacing and visual characteristics, it ignores the FFmpeg scene-detection output and estimates shot count from duration using an assumed three seconds per shot. Its BPM value is derived from that estimate, not detected from the soundtrack. The output is an analysis object and strategy-context text rather than an edited video. Local probing, temporary frames, and caching are supporting implementation details, not independent grounds for the mismatch.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The mismatch is a materially different primary purpose, not merely a supporting implementation detail. HitTextAnalyzer operates on text and produces analytical metadata and prompt context; there is no connection in this chunk to music timing or video-editing operations. Its local cache reads and writes support that text-analysis function. No network transmission or malicious behavior is evident. This assessment is limited to the supplied code chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code has a materially different primary purpose from the description. It parses CSV records and invokes hit_library.import_text_only or hit_library.import_video; it does not accept a song or implement the advertised beat-synchronized editing workflow. Although importing footage could support a larger video application, this chunk specifically manages a library of advertising examples, with no demonstrated connection to the declared editing capability. The internals of the imported library are not supplied, so no additional behavior is inferred.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied chunk serves advertising-performance analysis and training-data management rather than beat-synchronized video production. It reads CSV files, searches local directories for matching video paths, and queries or modifies analytics database tables. These are substantial capabilities not described by the declared video-editing purpose. The chunk contains no audio beat detection, video editing, text effects, reverse playback, or rendering. Although it could belong to a larger video application, the analytics and training functionality shown is materially distinct from ordinary implementation support for the declared task. No network transmission or malicious behavior is evident in this chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied chunk primarily performs historical content-performance analytics, not beat detection or automatic video editing. Although some analyzed factors concern music tempo and shot timing, the code does not ingest songs or footage, synchronize cuts, add flashing text, reverse clips, or render videos. Its repository-backed factor analysis is a materially different capability from the declared video-generation purpose. No malicious behavior or external data transmission is evident in this chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied chunk has a materially different primary function: generating spoken narration rather than editing visuals to musical beats. Its timing calculations concern speech segments and pauses, not music analysis or beat alignment. Although narration could support a broader video application, neither that feature nor its external speech-service processing is disclosed in the description. This assessment is limited to the supplied code and does not establish what other modules implement.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied chunk has a materially different purpose from the description: it is a local draft-discovery and style-import utility. It contains no music beat analysis, footage selection, reverse editing, flashing-text generation, or video rendering. Importing style could support a broader editing workflow, but automatically discovering and reading existing local projects is a distinct capability not disclosed by the description's song-and-footage workflow. No network transmission, file modification, or destructive behavior is shown.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied chunk's primary function is narration synthesis, not beat-synchronized video creation. It contains no music-beat analysis, visual asset selection, video cuts, flashing-text rendering, or reverse-video processing. Although narration could support a broader video workflow, this description does not mention narration or cloud speech synthesis, and the module introduces a distinct external text-processing capability. Credential use is ordinary authentication to the configured TTS provider, not evidence of credential theft. This assessment is limited to the supplied code chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code implements a script-diversity evaluator rather than beat-based video editing. It extracts structural signatures, compares scripts within a batch and against history, applies a passing threshold of 70, and generates revision suggestions. It contains no music analysis, beat detection, visual-asset selection, video cutting, text-overlay rendering, reverse playback, or video export. Although script evaluation could support a broader production pipeline, its advertising-copy diversity purpose is materially different from the declared music-driven editing functionality.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The primary workflow requires a ProductProfile and generates advertising strategies, narration scripts, and commercial video variants. In _build_shot_targets, shot boundaries are calculated from narration duration and a configurable fixed shot length (default 2500 ms); BGM is selected or rotated separately. The supplied code does not demonstrate beat detection, accent-based visual selection, flashing large-text effects, or reverse-play montage logic. Although both the description and implementation involve automated video editing, their primary purposes and timing mechanisms materially differ. This assessment is limited to the supplied chunk; implementations of delegated rendering components are not shown.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code implements an advertising-text compliance checker, a materially different purpose from automated beat-synchronized video editing. It contains no music analysis, beat detection, footage selection, video editing, or rendering functionality. Its local rule-file access supports the undeclared text-screening capability; no network access or malicious behavior is evident.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied module has a materially different purpose from the declared beat-synchronized video editor. It processes product selling points for advertising copy rather than analyzing music, detecting beats, editing footage, or applying video effects. Its local cache supports that unrelated marketing workflow. Although this is only one code chunk and does not establish what the full application does, the behavior shown is not a supporting implementation detail of the described video-editing functionality.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied module implements a talking-avatar subsystem, not beat-synchronized montage editing. Its avatar scanning and lip-sync generation are materially different capabilities that the description does not disclose. Although these could be optional features in a larger video pipeline, they are not ordinary implementation details of the described beat-cutting workflow. This chunk contains no beat detection, rhythm-based clip selection, flashing-text composition, or reverse-play editing. The bridge implementation is not supplied, so no additional behavior inside that subprocess is assumed.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code implements an advertising ideation matrix rather than music-driven video editing. It processes product categories, selling points, and target audiences; it does not process music or footage, detect beats, or perform video cuts and effects. The marketing-oriented behavior is materially different from the declared purpose, and the chunk provides no connection establishing it as a supporting component of beat-synchronized editing.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code implements an advertising-angle history and selection component, not beat detection or video editing. Its product-specific records, marketing metadata, 24-hour reuse cooldowns, and performance-based weighting indicate a materially different purpose. Local persistence alone is not the mismatch; the unrelated advertising-selection functionality is. No audio processing, footage manipulation, text effects, or video rendering appears in this chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied module's primary function is advertising script generation: its inputs are creative strategies and product profiles, and its outputs are structured storyboard objects. It does not accept music or footage or implement the beat-synchronized editing promised by the description. Although storyboard generation can support a broader video pipeline, this module specifically implements product-marketing copy generation, which is materially different from the declared music-and-footage editing workflow. This assessment is limited to the supplied code chunk; it does not establish what other modules may implement.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code has a materially different primary purpose from the description. It builds an advertising-strategy prompt, calls an LLM for structured strategy fields, and constructs a strategy object. It does not process music or visual assets, detect beats, synchronize cuts, apply text or reverse-playback effects, or render video. Local memory and campaign data support this advertising-strategy workflow, rather than the declared music-driven editing workflow. This assessment is limited to the supplied code chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied module is centered on product advertising and conversion-oriented copywriting. It contains no music analysis, beat detection, visual-asset processing, video editing, or rendering. Although script generation could support a broader video workflow, this chunk provides a materially different advertising-focused capability with no demonstrated connection to the declared beat-sync functionality. No external resource access or permission use is shown.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

该代码块的核心用途是广告文案生成模式门控,而非音乐节拍驱动的视频剪辑。它只进行条件判断并返回枚举或提示文本,没有节拍检测、视频素材处理、画面节奏对齐、快闪字幕或倒放操作。代码中的“混剪”指组合参考文案的开场、卖点和行动号召结构,并非声明的视频混剪。尽管这些逻辑可能被更大的视频制作系统调用,所提供代码呈现的主要职责与声明用途存在实质差异;未发现外部资源访问或权限使用。

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied module targets making near-identical video variants appear different to platform deduplication systems, rather than synchronizing edits to musical beats. Its executable behavior supports that purpose through subtle cropping, scaling, mirroring, frame-rate changes, and similarity checks. Although video transformation can support ordinary editing, this dedicated anti-deduplication capability is materially distinct from and absent from the declared description. The chunk does not implement the audio, OCR, or ASR modifications mentioned in its header, and it contains no evidence of network access or data exfiltration.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied module explicitly manages generated-copy diversity history. It contains no audio analysis, beat detection, media editing, text-overlay rendering, reverse playback, or video export. Although caching can support a larger application, this code's copy/script-signature and product-metadata focus is materially different from the declared music-video workflow. This assessment is limited to the supplied code chunk; it does not establish what the rest of the skill implements.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied chunk implements speech narration generation rather than music-beat detection or synchronized video editing. Narration could support a broader video application, but the declared workflow does not mention text input, generated speech, or sending text to a third-party synthesis service. These are materially distinct capabilities, not merely implementation details of beat-synchronized editing. This assessment is limited to the supplied chunk and does not establish whether other modules implement the advertised video features.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied chunk primarily implements an advertising asset and analytics library, not a beat-synchronized video editor. It contains no music beat detection, synchronization, montage rendering, flashing-text generation, or reverse-playback editing. Thumbnail generation and video-factor extraction could support a broader editing system, but the extensive marketing-copy, performance-metric, and prompt-reference functionality is materially different from the declared purpose. The code also initializes local storage on module import. No explicit network transfer or execution of generated prompt content is shown; the assessment is limited to this chunk.

Content

No source excerpt is available for this finding.