The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands to analyze local source code. This includes using rg (ripgrep) for pattern matching and cargo for building benchmarks and running tests within the Substrate development environment.
[EXTERNAL_DOWNLOADS]: The skill references and downloads tools from established repositories, such as the test-fuzz project from Trail of Bits and official FRAME documentation from Parity Technologies. These resources are well-known within the blockchain development community and are directly related to the skill's auditing purpose.
[INDIRECT_PROMPT_INJECTION]: The skill processes external source code provided by users, which creates an attack surface for indirect prompt injection. A malicious project could contain code or comments designed to manipulate the agent's reporting logic.
Ingestion points: The skill reads Rust source files (.rs) and configuration files (Cargo.toml) from the local workspace.
Boundary markers: No explicit delimiters are specified for the processed content, although the skill uses structured pattern matching for its analysis.
Capability inventory: The agent can execute file system searches using rg and build/test commands using cargo.
Sanitization: The skill relies on the agent's ability to interpret command outputs and code patterns to generate its vulnerability report.

substrate-vulnerability-scanner