The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted user input to populate generated code templates.
Ingestion points: User requirements collected during 'Phase 1: Requirements Gathering' (such as character bio, traits, and topics) are directly interpolated into TypeScript and Markdown templates.
Boundary markers: The skill uses template markers like {CharacterName} for interpolation but does not implement explicit delimiters or instructions in the resulting files to mitigate the risk of the agent interpreting embedded malicious instructions.
Capability inventory: The skill has the capability to perform file system operations using Write, Edit, and mkdir tools to establish the character project structure.
Sanitization: No sanitization, escaping, or validation logic is specified for the user-supplied strings before they are written to the disk.
[CREDENTIALS_UNSAFE]: The skill includes code blocks that generate .env.example files containing placeholders for sensitive secrets, including OPENAI_API_KEY, ANTHROPIC_API_KEY, and a database connection string with a hardcoded placeholder password (postgresql://user:pass@db-host:5432/eliza). These are identified as template examples rather than active secrets.
[EXTERNAL_DOWNLOADS]: The skill defines a package.json file that includes dependencies such as @elizaos/core, @elizaos/plugin-bootstrap, and @elizaos/plugin-sql. These are recognized as standard components of the well-known elizaOS framework ecosystem.

character-generator