The Agent Skills Directory

[SAFE]: The skill serves as a reference and scaffolding tool for plugin development. All identified command execution patterns (e.g., directory creation, git configuration reading, and file system management) are strictly limited to the primary purpose of initializing and managing plugin projects.
[DATA_EXPOSURE_SAFE]: While the skill accesses git config for user.name and user.email, this is done locally to pre-populate manifest metadata for new plugins and does not involve external exfiltration. The use of environment variable placeholders in documentation (e.g., STRIPE_API_KEY, DATABASE_URL) follows security best practices by discouraging hardcoding of secrets.
[REMOTE_CODE_EXECUTION_SAFE]: The skill mentions external Model Context Protocol (MCP) servers and tools like Prettier. These are referenced as examples of well-known, legitimate services (such as Stripe and official Model Context Protocol packages) used for extending plugin functionality. No unverified or piped remote execution (e.g., curl | bash) was found.
[PROMPT_INJECTION_SAFE]: The agent prompt templates provided in the examples use standard role-playing patterns for domain expertise (e.g., Docker expert, API designer) without attempting to bypass safety filters or override system instructions.
[NO_OBFUSCATION]: Analysis confirmed the absence of Base64 encoding, zero-width characters, homoglyphs, or other techniques intended to hide malicious intent.

plugin-master