The Agent Skills Directory

EXTERNAL_DOWNLOADS (MEDIUM): The skill mandates installing a global tool using uv tool install silverback. While silverback is a known framework by ApeWorX, it is not on the provided trusted source list. Additionally, it requires fetching external documentation from docs.apeworx.io to function.
COMMAND_EXECUTION (MEDIUM): The skill is designed to generate a bot.py file and execute it using silverback run. Running dynamically generated code that interacts with blockchain networks (signing transactions) is a high-risk operation.
PROMPT_INJECTION (LOW): The skill exhibits an indirect prompt injection surface (Category 8). It explicitly instructs the agent to use web_fetch on an external documentation site (docs.apeworx.io) to determine its logic.
Ingestion points: web_fetch from docs.apeworx.io inside SKILL.md.
Boundary markers: Absent; there are no instructions to ignore malicious commands embedded in the fetched documentation.
Capability inventory: The agent has the ability to install tools, write files, and execute shell commands.
Sanitization: Absent; the fetched content is used directly to guide code generation.
CREDENTIALS_UNSAFE (LOW): The skill encourages the use of environment variables for sensitive data like signers and API keys. While this is a standard practice, the skill also suggests print debugging, which could lead to accidental credential exposure in logs.

writing-bots