The Agent Skills Directory

[SAFE]: No malicious patterns, persistence mechanisms, or unauthorized data access were identified. The skill follows established security best practices.
[EXTERNAL_DOWNLOADS]: The skill fetches token metadata and risk reports from the official B20 Console API at https://b20.charon.codes/api/inspect. This is consistent with the skill's stated purpose of contract inspection.
[COMMAND_EXECUTION]: Executes a local Node.js script (scripts/inspect-b20.js) to interact with the inspection API. The script performs robust local validation of the contract address using a regular expression (0x[a-fA-F0-9]{40}) to ensure only valid Ethereum addresses are processed.
[PROMPT_INJECTION]: The skill manages a potential indirect prompt injection surface by processing data from a remote API.
Ingestion points: Contract metadata, policy labels, and risk flags are ingested from the B20 Console API endpoint.
Boundary markers: Explicit instructions in SKILL.md and references/api.md define a trust boundary, warning the agent that API responses are advisory and must not be treated as instructions for transaction execution or wallet configuration.
Capability inventory: The skill's operations are restricted to read-only network GET requests. It lacks capabilities for file system modification, environment variable access, or system-level privilege escalation.
Sanitization: The helper script implements a sanitize function that filters control characters and potential injection symbols (e.g., backticks, angle brackets). It also maps API risk and policy codes to a local allowlist of trusted descriptions.

b20-console