Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

PonyFlash - Media Generation Router

v1.0.0

Generate images, videos, speech audio, and music using the PonyFlash Python SDK. Also handle local media editing with FFmpeg, including clip, concat, transco...

1· 83·0 current·0 all-time
byPonyflash@leothebravest
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
!
Purpose & Capability
The skill claims both PonyFlash cloud generation and local FFmpeg editing — that mapping is coherent. However, the registry metadata declares no required environment variables or primary credential, while the SKILL.md explicitly requires a PONYFLASH_API_KEY for cloud tasks. That mismatch (undeclared API key requirement) is an incoherence worth flagging.
!
Instruction Scope
Runtime instructions ask the agent to: prompt the user for an API key (and suggest pasting it into chat), export it as PONYFLASH_API_KEY, pip install the 'ponyflash' package, run SDK calls to verify balance, run local shell scripts (check_ffmpeg.sh, media_ops.sh), and download subtitle fonts. Asking users to paste secrets into chat and instructing runtime package installation are both scope-expanding behaviors that require user caution.
Install Mechanism
The skill has no formal install spec, but SKILL.md tells the agent/user to run `pip install ponyflash` at runtime. That causes code to be pulled from PyPI (or another pip index) when executed. The included shell scripts also download fonts from two external URLs (mirrors.aliyun.com and jsdelivr). These are plausible for the use-case but increase attack surface compared to a purely instruction-only skill — review the 'ponyflash' package and the font sources before running.
!
Credentials
The registry lists no required env vars, yet SKILL.md and scripts use/expect: PONYFLASH_API_KEY (sensitive), PONYFLASH_FONT_DIR / PONYFLASH_NOTO_FONT_URL (optional), HOME and PATH. The skill also directs the user to paste the API key into chat. Requesting an API key is reasonable for a cloud SDK, but the missing declaration in metadata and the recommendation to paste the key in chat are both problematic from a credential-proportionality and secret-handling perspective.
Persistence & Privilege
The skill does not request always:true, does not modify other skills, and only writes local cache and temporary output files (e.g., ~/.cache/ponyflash/fonts/ and temp task dirs). This is expected for subtitle/font caching and media processing. No elevated or persistent platform privileges are requested.
What to consider before installing
What to consider before installing/using this skill: - Secret handling: The SKILL.md asks you to provide a PonyFlash API key (PONYFLASH_API_KEY). The registry metadata does not declare that key — treat this as a red flag and avoid pasting secrets directly into public chat logs. Prefer setting the API key in a secure per-agent secret store or as an environment variable scoped to the agent process, not in an open chat message. - Runtime installs: The instructions tell you to run `pip install ponyflash` at runtime. Before doing that, review the 'ponyflash' package on PyPI (or its source repository) to ensure it is legitimate and inspect its code. Consider installing in a fresh virtualenv or sandbox rather than system-wide. - External downloads: The included scripts download subtitle fonts from mirrors.aliyun.com and jsdelivr. Those are public CDNs; this behavior is plausible, but confirm you are comfortable with the network calls and the specific URLs (you can override them with PONYFLASH_NOTO_FONT_URL). - Local script behavior: The shell and Python scripts operate on local files and create caches under ~/.cache/ponyflash/fonts and temporary task directories. If you run media_ops.sh or ensure_subtitle_fonts.sh, they will read/write those locations. Run these scripts in a controlled workspace and inspect them first if you have privacy concerns. - How to reduce risk: - Do not paste API keys into chat. Configure PONYFLASH_API_KEY via your agent's secret/environment configuration or provide it interactively in a private, secure channel if your agent supports it. - Inspect the 'ponyflash' SDK source or pin a known-good version before installing. Use a virtualenv or container for runtime installs. - If you only need local FFmpeg editing, you can avoid the cloud path — the FFmpeg scripts do not require the API key. - If you must run the skill, run it in a sandbox or test environment first and verify network activity (which hosts the SDK communicates with) and the outputs before trusting it with sensitive data. - What would change this assessment: If the registry metadata is updated to explicitly declare required env vars (PONYFLASH_API_KEY) and primary credential, and if there is a clear, discoverable, trusted source for the 'ponyflash' SDK (official GitHub/PyPI with matching provenance), my confidence would increase toward 'benign'. Conversely, evidence of the 'ponyflash' pip package being untrusted/malicious would raise severity. Overall: the skill appears to implement the advertised features, but the undeclared API-key dependency, the recommendation to paste the key into chat, and runtime package downloads are coherence and operational-risk issues — proceed with caution.

Like a lobster shell, security has layers — review code before you run it.

latestvk977eewwchngbyemnmr5g7h3e583ghg1

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Comments