The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/post.mjs contains a hardcoded API key quack_5c6786fb7ac8f08c638d5db176d60b12b906b78ac63d21cf. Hardcoding secrets in source code is a high-risk practice that exposes credentials to any user with access to the skill files, potentially leading to unauthorized account access.
[DATA_EXFILTRATION]: The skill uses the fetch API in scripts/post.mjs to send user-provided text to a remote endpoint at https://vibepost-jpaulgrayson.replit.app/api/quack/post. This creates a channel where sensitive data could be sent to an external server if the agent is manipulated into including it in the post content.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted text input and has the capability to publish it externally. The SAFETY.md file identifies this surface and attempts to mitigate it with behavioral instructions.
- Ingestion points: CLI --text argument in scripts/post.mjs used for the social media post content.
- Boundary markers: Present in SKILL.md and SAFETY.md as explicit instructions to the agent to confirm text and refuse reading local files for content.
- Capability inventory: Network POST capability via fetch in scripts/post.mjs to an external API.
- Sanitization: Absent; the script directly transmits the provided string without validation or escaping.
[COMMAND_EXECUTION]: The skill executes a local Node.js script scripts/post.mjs which performs network requests. This provides the primary mechanism for the agent to transmit data to external services.

social-poster-hardened