The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a variety of shell commands to perform its auditing functions. This includes running the euthyna.js CLI tool via node, analyzing git history with git blame, and invoking build/test tools such as go test, cargo test, and pytest. While these are standard auditing operations, they involve the execution of scripts and binaries on the host system.- [DYNAMIC_EXECUTION]: The skill provides instructions for running proof-of-concept (PoC) reproduction scripts and project-specific test suites to gather coverage data. This requires executing code from the repository being audited, which poses a risk if that code is malicious. The skill mitigates this by requiring explicit user approval and recommending sandboxed environments.- [INDIRECT_PROMPT_INJECTION]: The skill audits external source code, pull requests, and security reports, which are untrusted data sources. Malicious actors could potentially embed instructions or misleading data in these inputs to subvert the agent's analysis. The skill implements a rigorous 'gate' system (Reachability, Impact, PoC Verification, etc.) and a 'devil's advocate' review process specifically designed to prevent being misled by external claims.

euthyna-en