ClawHub

Bilibili AI Video Prompt Analyzer

v1.1.1

Analyze Bilibili AI-generated videos to extract/reverse-engineer prompts for imitation and replication. Extracts frames, extracts subtitles, generates detail...

⭐ 0· 114·0 current·0 all-time
by@jilanfang
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
Name/description ask for reverse-engineering prompts from Bilibili AI videos; required tools (yt-dlp, ffmpeg, ffprobe) and the three provided scripts directly support that purpose and are proportionate.
ℹ
Instruction Scope
SKILL.md and scripts limit actions to downloading the provided URL, extracting frames and subtitles, and producing text prompts. The skill asks for 'multimodal LLM support' (i.e., a vision-capable model) but does not declare or request any external API keys — analysis will rely on the agent's model. One operational note: download script uses --yes-playlist, which may download whole playlists if given a playlist URL (not strictly a security issue but could be broader than the user intended). Hardsub OCR is explicitly not implemented.
✓
Install Mechanism
Instruction-only skill with no install spec; scripts are small shell files (no archived downloads or third-party install URLs). This is low-risk from an install perspective.
✓
Credentials
No environment variables, credentials, or config paths are requested. The scripts only operate on the user-supplied video path and output directories — there are no requests for unrelated secrets or system config.
✓
Persistence & Privilege
always is false and the skill does not attempt to modify agent/system configuration or persist credentials. It writes only to the provided output directories.
Assessment
The skill appears coherent and the scripts are readable; still review before use. Practical points: ensure ffmpeg/ffprobe and yt-dlp are installed and up-to-date; be aware yt-dlp --yes-playlist can download more than a single video if you pass a playlist URL; hardsub OCR is not implemented so embedded burned subtitles won't be extracted by this version; the analysis step requires a multimodal (vision-capable) model — the skill does not supply or request API keys, so confirm your agent/model supports image inputs. Also consider legal and ethical issues: extracting prompts to imitate others' AI-generated work may have copyright or terms-of-service implications. If you have any doubts, inspect and run the scripts in a sandboxed environment first.

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

latestvk971y1y2621pnfhtt7sndrgctd834gn5

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments