The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to install the ACP client by cloning a repository from 'github.com/Virtual-Protocol/openclaw-acp' and running 'npm install'. This is documented as a prerequisite for interacting with the service platform and is treated as a reference to a well-known infrastructure service.
[COMMAND_EXECUTION]: The documentation provides multiple shell script examples utilizing system binaries such as 'npm', 'curl', 'jq', and 'bc', as well as local tools like 'acp' and 'solana-cli' to automate trading and portfolio risk assessment.
[DATA_EXFILTRATION]: The skill includes explicit security guidelines advising agents and users to remove private keys, wallet seeds, and API credentials before submitting token addresses for verification, which promotes safe data handling.
[PROMPT_INJECTION]: The skill's primary function is to process external data (Solana contract addresses) and provide a security audit that drives automated agent decisions. This constitutes an indirect prompt injection surface where a service response could influence agent logic.
Ingestion points: Solana contract addresses submitted via the 'acp job create' requirements (SKILL.md).
Boundary markers: Audit findings are returned in a structured JSON schema, providing clear separation between data fields.
Capability inventory: The skill is designed for integration into scripts with capabilities to execute trades ('solana-cli trade') and manage portfolios based on external assessment results.
Sanitization: Example scripts leverage 'jq' for parsing structured output, although the final decision-making logic depends on the high-level 'risk_score' and 'safe_to_trade' indicators provided by the external service.

cybercentry-solana-token-verification