The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted Solidity source code, including NatSpec and inline comments, which are summarized by subagents for the final report. 1. Ingestion points: Project source files are read during Step 2 of the pipeline using Read and Glob tools. 2. Boundary markers: The agent uses structured extraction prompts with specific headers to delimit code content. 3. Capability inventory: The skill has file-write access and can execute shell commands for report and diagram generation. 4. Sanitization: Internal scripts perform basic formatting and comment stripping for structural analysis, but subagents receive larger blocks of content for summarization.
[COMMAND_EXECUTION]: The skill executes local scripts provided in the skill directory, such as analyze_git_security.py and entry_points.py, and invokes forge and hardhat for coverage analysis. These operations are restricted to the local environment and the analyzed project root.
[EXTERNAL_DOWNLOADS]: Fetches a version string from the author's public GitHub repository to perform a version check. This is a neutral administrative operation from a vendor-owned resource.

x-ray