The Agent Skills Directory

[SAFE]: The installation script scripts/bootstrap-env.sh downloads binaries for the wallet CLI and TSS node from verified vendor-owned domains (cobo.com). These operations are consistent with the skill's primary purpose and utilize established service endpoints.
[SAFE]: The prompt injection patterns identified in references/security.md are documented as negative examples within a defensive guide. These instructions explicitly teach the agent to refuse and report malicious attempts found in external content.
[SAFE]: The skill uses official package managers (pip, npm) to install vendor-maintained libraries (cobo-agentic-wallet, @cobo/agentic-wallet).
[SAFE]: Dynamic script execution and management in references/sdk-scripting.md are core features intended for wallet automation and strategy execution. The instructions emphasize local storage, parameterization, and script reuse rather than execution of untrusted remote content.
[SAFE]: Transaction security is reinforced by a mandatory 'Execution Authorization' flow, ensuring that high-risk operations like smart contract calls require manual approval from the wallet owner.

cobo-agentic-wallet-sandbox-test