The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8c: Tool output poisoning). It instructs the agent to fetch and process info.json metadata from the trustwallet/assets GitHub repository. Since this repository accepts public contributions via Pull Requests, an attacker could potentially embed malicious instructions within token descriptions or other metadata fields that the agent might subsequently follow.
Ingestion points: Reads token metadata using gh api from the trustwallet/assets repository as described in SKILL.md.
Boundary markers: No delimiters or "ignore embedded instructions" warnings are specified for the fetched JSON content.
Capability inventory: The skill references execution of local make commands and gh api calls in SKILL.md.
Sanitization: No sanitization or validation of the external JSON content is mentioned before it is processed by the agent.
[COMMAND_EXECUTION]: The skill includes several make commands (e.g., make check, make fix, make add-token) intended for local asset management and validation. While these are standard project maintenance scripts, they represent a capability to execute local commands on the host system if the repository is cloned.

assets