The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill outlines architectures that ingest untrusted external data which is then used to perform onchain actions, creating a potential indirect prompt injection surface.
Ingestion points: Untrusted data enters the system through onchain_event triggers or Turbo Webhook sinks (SKILL.md).
Boundary markers: No specific prompt boundary markers are prescribed for the agent when handling this data.
Capability inventory: The system uses wallet.writeContract and sendTransaction for onchain execution (SKILL.md).
Sanitization: The skill provides proactive security guidance, instructing the developer to "decode and validate before acting" on incoming webhook bodies (SKILL.md).
[SAFE]: The instruction to correct technical misconceptions about Goldsky's execution capabilities is a matter of factual accuracy for the product's scope and is not an attempt to conceal malicious activity or bypass safety filters.

onchain-automation