The Agent Skills Directory

[SAFE]: The skill implements strong security guardrails by repeatedly instructing the agent to never display or request the values of secrets, tokens, or credentials. It specifically forbids common insecure debugging practices such as printing process.env or disabling TLS verification.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect injection as it is designed to process external data such as application logs and user-provided error symptoms, but this risk is well-mitigated by the provided instructions.
Ingestion points: The skill ingests untrusted data from project configuration files, application server logs, and user-reported error messages via the Read and Grep tools.
Boundary markers: Explicit instructions are provided to redact user data and credentials, and to stop processing logs if they appear to contain customer content or secrets.
Capability inventory: Capabilities are restricted to read-only file access (Read) and pattern searching (Grep). No network exfiltration or command execution tools are authorized in the frontmatter.
Sanitization: The instructions require the agent to manually redact tokens, credentials, and PII from its output, ensuring sensitive data processed from the environment is not exposed in the conversation history.
[EXTERNAL_DOWNLOADS]: The skill references resources from Replit's official documentation and status domains (docs.replit.com, status.replit.com). As a well-known service provider relevant to the skill's primary purpose, these references are documented neutrally as legitimate resources.

replit-common-errors