C — Moderate risk signals Graded October 16, 2026Several concerns worth a closer look before you install.Based on public adoption and transparency signals as of October 16, 2026. Our opinion, not a code audit or security guarantee.
AI agent skill · Developer tools, Design & media
lemo-opuscar
Claude Code skill for short films with no video model: 51 film styles, each a style prompt plus a demo film made entirely in code by Claude Opus 5.5 (incl. OPUSCAR 98 and the new Evidence Films). By Lemomo. · Opus5.5 x 51 种影片风格：风格提示词 + 纯代码样片 + 导演与技术指南（新增「实证影像」）
Bottom line: Signal-based grade (3/7 points): well starred on GitHub, public source code. This reflects adoption and transparency, not a code audit.
- Popularity score
- 39/100
- Rank
- Not ranked: outside the 500 most popular tools
- GitHub stars
- 2.5k
What it does
Lemo-opuscar is a Claude Code skill that generates short films in 51 different visual styles using Claude Opus 5.5, with each style accompanied by a style prompt and a demo film created entirely through code. It includes cinematic styles like OPUSCAR 98 and emerging Evidence Films, along with director and technical guides.
Who it’s for
Creative directors, filmmakers, and builders exploring AI-driven visual storytelling who want to experiment with diverse cinematic styles without traditional video production. Best suited for prototyping, concept visualization, and understanding Claude's code execution capabilities.
Copy-paste setup
Check these steps against the project’s own instructions before running anything.
Verify against the repo README at https://github.com/lemomo-ai/lemo-opuscar (opens external site) for the most current setup instructions. Likely usage involves cloning the repository and invoking Claude with the provided style prompts; check for Claude Desktop integration or direct API usage patterns in the documentation.
Risks
- Code execution risk: Claude generates and may execute code to render films; malformed or adversarially-crafted style prompts could cause performance degradation or unexpected behavior.
- Prompt injection: Style prompts are user-controlled; insufficient input validation could allow injection of arbitrary instructions to Claude.
- Resource consumption: Generating multiple film styles on-demand may incur high Claude API costs; no apparent rate-limiting or quota documentation visible.
- Output licensing unclear: No explicit terms on generated film ownership or redistribution rights; potential IP ambiguity if films are commercialized.
- Supply chain maturity: 2.5K GitHub stars but no visible npm package presence suggests early-stage tooling with limited production hardening.
How to stay secure
Security Hardening Guide: lemo-opuscar
Least-privilege permissions
- Restrict API access: Only grant the Claude API key permissions for film generation tasks; do not share credentials or allow the skill to access other cloud services, databases, or file systems.
- Sandbox code execution (technical term: running code in an isolated container): Run film rendering in a separate, restricted environment that cannot modify your main systems or access sensitive files.
- Separate user roles: Ensure only approved team members can create film style prompts; prevent general users from directly writing or editing the prompts that feed into Claude.
Network isolation
- Limit outbound connections: Configure firewalls to allow the skill only to communicate with Claude's API servers; block any unexpected outbound requests to other services.
- No direct internet access for rendering: Keep the film generation process offline from public networks when possible, or use a dedicated, monitored connection for API calls only.
- Audit log access: Track all API calls and data sent to Claude; maintain records of who requested what films and when.
Monitoring
- Set usage alerts: Monitor Claude API costs daily; configure automatic warnings if spending exceeds your expected budget to catch runaway requests early.
- Track prompt submissions: Log every style prompt entered into the system; flag unusual requests or repeated attempts with similar suspicious patterns.
- Performance baselines: Establish normal response times and resource usage; alert if rendering suddenly becomes much slower (sign of malicious complexity injection).
Red flags to watch for
- Unusual style prompt requests: Watch for prompts containing code syntax, system commands, or instructions that seem unrelated to filmmaking—these may be injection attempts.
- Cost spikes: A sudden increase in API bills without corresponding increase in film requests suggests adversarial or accidental overuse.
- Generated films behaving unexpectedly: If rendered films crash viewing tools, consume excessive memory, or display unintended content, immediately disable the skill and review the prompt that created them.
- Unclear licensing claims: Before commercializing any film, verify in writing who owns it (you, Lemomo, or Claude/Anthropic); do not assume free redistribution rights.
Alternatives
- Runway ML – full video generation API with built-in style controls and managed API security.
- Synthesia – text-to-video platform with templated cinematic styles and clearer licensing.
- Pika Labs – AI video generation with style library; more polished UX but closed-source.
- Direct Claude Vision + image APIs – manual style prompting without bundled skill layer, lower abstraction.