The Agent Skills Directory

[METADATA_POISONING]: The skill metadata field homepage contains the URL https://github.com/tedkaczynski-the-bot/game-theory. The account name tedkaczynski-the-bot references a domestic terrorist and is highly inappropriate for professional skill metadata. Furthermore, this name is inconsistent with the stated author sundial-org, indicating deceptive or misleading metadata placement.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze complex external data (protocol tokenomics, governance proposals, MEV strategies) which presents an attack surface for instructions hidden within that data.
Ingestion points: The SKILL.md instructions and various reference documents (e.g., governance-attacks.md, tokenomics-analysis.md) prompt the agent to analyze external protocol specifications and community data.
Boundary markers: No boundary markers or "ignore embedded instructions" warnings are included in the Analysis Template or methodology sections to protect the agent from adversarial content in the analyzed protocols.
Capability inventory: The skill does not define dangerous tools such as shell access or network operations, limiting the impact of a potential injection to context poisoning.
Sanitization: The methodology lacks any validation or sanitization steps for external data before it is processed by the agent.

game-theory