The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external financial data from third-party APIs, which represents a potential surface for indirect prompt injection. The risk is minimized by the skill's analytical focus and explicit lack of execution capabilities.
Ingestion points: The skill ingests wallet positions and protocol data from external providers including Zerion, DeBank, CoinGecko, and DefiLlama as documented in references/providers.md.
Boundary markers: The instructions mandate clear labeling of data sources, observation timestamps, and coverage gaps in all generated reports.
Capability inventory: The skill is strictly limited to data aggregation and reporting. It explicitly disclaims any authority for fund movement, custody, or transaction execution in SKILL.md and references/reporting.md.
Sanitization: The instructions emphasize rigorous, idempotent reconciliation logic to process raw balances and debt figures before they are presented in the final portfolio report.
[SAFE]: The skill incorporates security best practices by directing the agent to use secure credential stores for API keys and strictly forbidding the logging of authentication headers in references/providers.md.
[SAFE]: The skill maintains a clear separation between reporting and action, stating that proposed rebalance plans are estimates that do not constitute authorization for execution or account modification.

galleon-defi-portfolio