Automate TikTok slideshow marketing for any app or product. Researches competitors, generates AI images, adds text overlays, posts via Postiz, tracks analyti...
Security Analysis
medium confidenceThe skill's code and instructions match its stated TikTok marketing purpose, but the registry metadata omits the many credentials/config paths the scripts actually need and the skill will store/use sensitive API keys — review config handling and run in a sandbox before trusting it.
The skill is clearly a TikTok slideshow marketing pipeline and the scripts legitimately require Postiz and an image-generation provider plus optional RevenueCat access — this aligns with the stated purpose. However, the registry metadata claims 'no required env vars / config paths' while SKILL.md and the scripts expect a config (e.g., config.postiz.apiKey, config.imageGen.apiKey, revenuecat keys) and external accounts (Postiz, OpenAI/Stability/Replicate, RevenueCat). That mismatch is incoherent and could mislead users about the secrets they must provide.
SKILL.md instructs the agent to browse competitor content (with user permission), generate images, overlay text, post via Postiz and correlate analytics with RevenueCat. The instructions and scripts operate on local files (config JSON, snapshots, reports) and call external APIs (Postiz, OpenAI/Stability/Replicate, RevenueCat). They ask for browser permission before scraping and advise manual steps for adding trending audio. The scope stays within marketing/analytics; there are no instructions to read unrelated system files, harvest arbitrary secrets, or send data to unknown endpoints beyond the documented services.
No install spec (instruction-only) — lower platform-level risk, but the skill requires Node.js and native dependency node-canvas which may require system build tools (Python, make, C++ compiler). The scripts will be run locally and will install dependencies via npm as needed; this is expected but users should be prepared to install native toolchains or run inside a container.
Functionality legitimately requires API credentials (Postiz API key, image-generation API key, optional RevenueCat secret keys). The problem is the package metadata declares 'none' for required env/config paths, yet the code expects a config.json and secret keys and will store snapshots/webhook logs locally. Requesting a RevenueCat V1 secret (sk_...) is particularly sensitive. The set of required secrets is proportional to the feature set, but the metadata omission and lack of explicit guidance about where/how credentials are stored is a security concern.
The skill does not request always:true and does not attempt to change other skills or system-wide agent settings. It writes files within its own workspace (reports, snapshots, competitor JSONs). Autonomous invocation is allowed (platform default) but not combined with 'always' or other privilege escalation in the codebase.
Guidance
What to consider before installing/using 'Larry': - Metadata mismatch: The registry shows no required env vars or config paths, but the SKILL.md and scripts require a config file containing Postiz API keys, image-generation API keys (OpenAI/Stability/Replicate), and optional RevenueCat secret keys. Treat that as an inconsistency — ask the publisher to update metadata or inspect the config format before providing secrets. - Secrets handling: The scripts expect and will use secret API keys (including RevenueCat V1 secret). Provide only least-privilege keys, prefer short-lived or scoped tokens where available, and avoid putting long-lived production secrets in an unencrypted file. Consider creating separate test accounts for Postiz and RevenueCat. - Local storage and I/O: The skill writes analytics-snapshot.json, platform-stats.json, rc-snapshot.json, hook-performance.json, competitor-research.json and report files. These may contain PII or business-sensitive metrics. Run the skill in an isolated directory and/or sandbox (container or VM) so data is contained. - Build requirements: node-canvas is a native module that often requires Python, make, and a C++ compiler. If you run this on your machine, ensure you understand and approve any system-level packages the agent will install, or run inside a prebuilt image. - Third-party dependencies & links: The SKILL.md encourages signing up at a specific Postiz referral link and installing a RevenueCat skill via ClaWHub. Verify those services and the referral are acceptable. Installing other skills can expand privileges; review them too. - Code review: The included scripts are readable and show the network endpoints used (Postiz, OpenAI/Stability/Replicate, RevenueCat). Before running, scan the actual onboarding and other scripts (onboarding.js not shown) for any steps that might transmit other local data or auto-run commands. Ask the author to disclose exactly where config files live and whether API keys are ever transmitted to other endpoints. - Operational recommendations: If you want to try it, run it in a disposable container or non-production environment, use test API keys, and inspect all generated files before providing production credentials. If you plan to provide RevenueCat secrets, limit scope or use a separate RevenueCat project for testing. If you want, I can: (1) extract a checklist of exact config keys/paths the scripts expect, (2) highlight any specific lines that read/write credentials, or (3) produce a safe run plan (container commands and minimal permissions) you can follow.
Latest Release
v1.0.0
Initial release
More by @OllieWazza
Published by @OllieWazza on ClawHub
