The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains text discussing prompt injection risks and how to mitigate them (e.g., in references/agent-security.md). It also includes prompt templates for users to give to their coding agents (e.g., in references/dev-methodology.md). These are instructional examples and do not attempt to override the primary agent's instructions.
[EXTERNAL_DOWNLOADS]: The skill references several npm and Python packages from official registries and provides a download link for an SDK hosted on a verified vendor domain (app.agentvault.io). It also references cloning or fetching metadata from official Celo GitHub repositories. These are legitimate resources for developers in the Celo ecosystem.
[INDIRECT_PROMPT_INJECTION]: The skill describes functionality for retrieving data from external APIs such as DefiLlama, The Grid, and Celo's governance APIs. While this constitutes an ingestion surface for untrusted data, the skill includes a dedicated security reference (references/agent-security.md) that explicitly warns developers about indirect injection and provides mitigation strategies like output escaping and input classification.
[DATA_EXFILTRATION]: No evidence of unauthorized data harvesting or sensitive file access. The skill provides standard guidance on credential management using .env files and configuration directories, following industry best practices.

celopedia-skill