The Agent Skills Directory

[SAFE]: The skill implements proactive mitigations for potential indirect prompt injection from the public search corpus. Ingestion of untrusted data from the Scry API is governed by Rule 11, which provides a sanitization mechanism using a content_risk filter. This prevents adversarial content from entering the agent's context through search results.
[EXTERNAL_DOWNLOADS]: Trial API keys are retrieved from the vendor's official endpoint at api.scry.io. This is an intended and documented feature for trial access and occurs within the vendor's own infrastructure.
[REMOTE_CODE_EXECUTION]: An automated scanner alert for a curl | python3 pattern was evaluated. The command uses Python's standard library to parse JSON data (python3 -c 'import json,sys; ...') from the vendor's API response. This is a standard utility pattern for data extraction and does not constitute the execution of untrusted remote code.
[COMMAND_EXECUTION]: The skill includes instructions for running a cargo run command to perform a contract audit. This is a development-oriented task that operates on local source code and utilizes official package registries for dependencies.
[SAFE]: Secret management follows secure industry standards. The skill recommends storing credentials in a .env file rather than hardcoding them or passing them through insecure channels.

scry