dotnet-docfx-digest

Pass

Audited by Gen Agent Trust Hub on Sep 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from target repositories, including source code, XML comments, and Markdown files, to generate API documentation and usage examples.
  • Ingestion points: The skill reads files such as .cs, .md, and .yml from target repositories during the documentation audit process.
  • Capability inventory: The skill possesses powerful capabilities such as executing shell commands via dotnet run (for scripts and sample compilation), performing network searches via gh, and writing documentation to the file system.
  • Boundary markers: There are no explicit markers or instructions used to delimit untrusted repository content from documentation commands.
  • Sanitization: No explicit sanitization, validation, or filtering of the repository-extracted content is mentioned before it is used to generate documentation or samples.
  • [PROMPT_INJECTION]: The skill contains 'Autonomous Persistence' and 'Anti-Fatigue' sections in its instructions that provide strong directives for the agent to bypass standard agent behaviors. These include instructions to 'actively override' typical pauses, 'do not ask for permission,' and 'override conversational habits.' These steering commands are designed to circumvent user review and interaction guardrails in favor of high-autonomy execution.
  • [DYNAMIC_EXECUTION]: The skill generates C# code fences based on repository evidence and then compiles these samples using the docfx.cs validator tool in isolated temporary projects. This runtime compilation of generated code constitutes a dynamic execution path for content derived from external repository data.
  • [COMMAND_EXECUTION]: The skill utilizes standard developer CLI tools, including dotnet, git, and gh, to perform its tasks. While these tools are essential for documentation automation, they provide an execution vector for instructions that could be embedded in the repository data the skill processes.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 30, 2026, 04:36 PM