Influencer

Create, manage, and scale AI-generated virtual influencers with consistent characters, multi-platform content, and monetization workflows.

Security Scan

The instructions match an AI-influencer content workflow, but the skill omits declared credentials and explicitly advises detection-evasion techniques — those inconsistencies and policy-evasion instructions are concerning and need clarification before use.

Purpose & Capability

The skill's stated purpose (create/manage virtual influencers) legitimately requires many third‑party services (TTS, image/video generators, face-swap, hosting) and associated credentials. However, the registry metadata lists no required environment variables or primary credential. The files (voice.md, video-gen.md, image-gen.md, etc.) reference services that require API keys (ElevenLabs, HeyGen, D-ID, Runway, Nano Banana Pro, etc.) and voice IDs; those are not declared in the skill metadata — an incoherence between claimed capability and the declared requirements.

Instruction Scope

The SKILL.md and companion files instruct the agent to create and manage content in ~/influencers/ (reading/writing identity.md, reference images, tools.md, analytics.md) which is consistent with a content workflow but grants broad file-system access to persona data. The instructions include explicit examples that reference environment variables and API calls (e.g., curl with $ELEVENLABS_API_KEY in voice.md). Notably, the guidance includes an 'Avoiding Detection' section that teaches how to reduce AI‑generated content detection — this is an active evasion strategy and elevates risk. The instructions also recommend face-swap and LoRA training workflows that, if misapplied, enable deepfakes; the compliance notes mention consent but the operational instructions leave user-supplied consent/controls to the operator.

Install Mechanism

This is an instruction‑only skill with no install spec and no code files. That keeps the on-disk footprint small and avoids arbitrary downloads or installations, which is low risk from an install-mechanism perspective.

Credentials

Although the registry shows 'Required env vars: none', the documentation contains concrete API usage that requires secrets (e.g., ELEVENLABS_API_KEY, likely keys for HeyGen/D‑ID/Runway/Nano Banana Pro). The skill also instructs the agent to persist tool credentials/voice_id in tools.md. Requiring multiple service keys without declaring them is disproportionate and unclear — it both hides the exact credential needs and could lead to ad-hoc credential placement (files or env) with unclear protections.

Persistence & Privilege

The skill does not request 'always: true' and is user-invocable (normal). It instructs creating a ~/influencers/ workspace and storing tokens/voice_ids there (tools.md). That is consistent for a content tool, but because the skill includes explicit evasion techniques and workflows that could be used to create deepfakes, autonomous invocation combined with later-provided credentials would increase blast radius. Recommend restricting autonomous runs until credentials and intent are validated.

What to consider before installing

This skill provides a complete playbook for producing and managing AI-generated influencer content, including how to use third-party APIs and how to organize persona files. Key things to check before installing or giving it credentials: 1) Ask the publisher to list exactly which API keys and environment variables the skill requires (ElevenLabs, HeyGen, D‑ID, Runway, etc.) and where/how they will be stored. 2) Do NOT provide real-person images or private data unless you have explicit, documented consent — the workflows enable face‑swap/deepfake creation. 3) The guide contains an 'Avoiding Detection' section (tells you how to make AI content less detectable); consider whether you want a tool that promotes evading detection and confirm you will comply with platform laws/TOS. 4) If you proceed, restrict the skill's runtime permissions: disable autonomous invocation or limit it to read/write only within a tightly scoped directory, and avoid storing API keys in plaintext files. 5) Verify licensing for all third‑party tools and music, and consult legal counsel for commercial/monetized use. If the publisher cannot clarify required credentials and data flows, treat the skill as risky and avoid providing secrets or enabling autonomous operation.

Workspace Structure

Each influencer lives in a dedicated folder:

~/influencers/
├── {persona-slug}/
│   ├── identity.md        # Name, niche, voice, personality
│   ├── reference/         # Base images for consistency
│   │   ├── face-ref-1.png
│   │   └── style-guide.md
│   ├── content/
│   │   ├── photos/        # Generated images by date
│   │   └── videos/        # Generated videos by date
│   ├── captions.md        # Caption templates, hashtags
│   ├── schedule.md        # Posting schedule
│   └── analytics.md       # Performance tracking
└── tools.md               # Configured generation tools

---

Quick Reference

Task	Load
Create new persona (identity, niche, aesthetics)	persona.md
Generate consistent photos	image-gen.md
Generate videos (talking head, lifestyle)	video-gen.md
Voice and audio (TTS, voice cloning)	voice.md
Content strategy and captions	content.md
Platform optimization (IG, TikTok, YT)	platforms.md
Monetization (brand deals, affiliates)	monetization.md
Legal and disclosure requirements	compliance.md

---

Persona Creation Checklist

Before generating any content:

• Define niche (fitness, lifestyle, tech, fashion, etc.)
• Create identity document (name, age, location, backstory)
• Generate 5-10 reference images for face consistency
• Define visual style (lighting, colors, settings)
• Create voice profile (if using TTS/videos)
• Draft personality guidelines for captions

---

Character Consistency Rules

Maintaining the same face/body across all content is CRITICAL.

For photos:

1. Generate base reference set first (5-10 images, multiple angles)
2. Use IP-Adapter or InstantID for every generation
3. Same seed + similar prompt structure = more consistency
4. Quality check EVERY image before posting

For videos:

1. Use face-swap on real footage OR
2. Generate with character LoRA if trained OR
3. Talking head tools (HeyGen, D-ID) with reference image

---

Content Generation Flow

1. Plan — Check schedule.md for what's needed today
2. Generate — Use appropriate tool (see image-gen.md or video-gen.md)
3. Review — Verify character consistency, quality
4. Caption — Write engaging copy matching persona voice
5. Schedule — Queue for optimal posting time
6. Track — Log in analytics.md after posting

---

Common Patterns

User says	Agent does
"Create new influencer"	Run persona creation flow, set up workspace
"Generate photos for today"	Check schedule, generate with consistency refs
"Make a TikTok video"	Generate 9:16 video with talking head or lifestyle footage
"Write captions for these"	Draft captions matching persona voice + niche hashtags
"How is she performing?"	Summarize analytics.md, suggest improvements
"Add brand deal content"	Generate sponsored content with disclosure

---

Tool Configuration

Store active tools in ~/influencers/tools.md:

## Active Tools
- Image: Nano Banana Pro (Gemini)
- Video: Kling / Runway
- Voice: ElevenLabs (voice_id: xxx)
- Lip Sync: HeyGen

Update when switching providers. All generation scripts read from here.