The Agent Skills Directory

[Prompt Injection] (SAFE): No instructions attempting to override agent behavior, bypass safety filters, or extract system prompts were detected. The documentation focus is purely on API implementation and migration.
[Data Exposure & Exfiltration] (SAFE): The templates correctly use environment variables (process.env.GEMINI_API_KEY) and Cloudflare Worker secrets (env.GEMINI_API_KEY) for authentication. No hardcoded secrets, sensitive file paths, or unauthorized network exfiltration were found. Network calls are restricted to the official Google API endpoint (generativelanguage.googleapis.com).
[Obfuscation] (SAFE): No multi-layer encoding, zero-width characters, homoglyphs, or hidden executable commands were found in the source code or documentation.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill uses standard, well-known dependencies from the npm registry, specifically the official @google/genai SDK. No remote scripts are downloaded or executed at runtime. The check-versions.sh script is a local utility for package verification with no network-to-shell execution patterns.
[Privilege Escalation] (SAFE): No commands for privilege escalation (e.g., sudo, chmod 777, or system configuration modification) were detected.
[Persistence Mechanisms] (SAFE): No attempts to establish persistence through shell profiles, cron jobs, or startup services were found.
[Metadata Poisoning] (SAFE): The plugin metadata in plugin.json and the markdown headers are accurate and descriptive of the skill's actual functionality.
[Indirect Prompt Injection] (LOW): As a chatbot template, the skill naturally has ingestion points for untrusted data (e.g., in cloudflare-worker.ts via request.json()). While it has the capability to perform network operations (API calls), it does not provide downstream write access to sensitive systems. This is a standard risk for any LLM-based application template and is managed by the underlying LLM's safety filters.
[Time-Delayed / Conditional Attacks] (SAFE): No logic gating behavior based on time, date, or specific environmental triggers was detected.
[Dynamic Execution] (SAFE): The skill demonstrates the codeExecution tool feature of the Gemini API. This allows the AI model to run generated Python code within a Google-managed sandbox. This is a primary intended feature of the API and does not represent a vulnerability in the skill's own code.

google-gemini-api