Inline Comment

Best way to steer your agents, effortlessly.

Install
cmdop skills install agensi-inline-comment

Inline Comment is a skill published by agensi that gives developers a browser-based editor for leaving anchored feedback on code, configuration files, and documentation. Instead of quoting snippets or citing line numbers in chat, a developer selects the exact lines or code spans they want to address and writes a comment there. When submitted, the skill sends the AI a structured JSON object in which each comment is mapped precisely to the selected region, so the agent receives unambiguous instructions without requiring extra context-copying.

The editor supports syntax highlighting for over 80 languages and can render Markdown, KaTeX math expressions, and Mermaid diagrams. Code folding lets reviewers collapse sections of long files to navigate efficiently. The interface is designed with touch-sized targets and a responsive layout, so reviews can be submitted from a phone or tablet. It also operates over SSH port forwarding or HTTP tunnels, making it usable when working on remote machines through a local browser. Draft comments are saved automatically in the browser, so an accidentally closed tab does not discard in-progress work.

Supported file types include source code in languages such as Python, JavaScript, Rust, and Go; configuration formats like YAML, JSON, and Dockerfile; and documentation formats including Markdown and LaTeX. This skill is appropriate when an agent workflow requires precise, location-specific human feedback on text or code, rather than free-form conversational instruction.

Use cases

  • Leave line-specific refactoring instructions on a Python file without quoting code in chat
  • Annotate a YAML configuration file with targeted change requests before the agent applies them
  • Review AI-generated Markdown documentation and attach per-paragraph revision notes
  • Provide anchored feedback on a Dockerfile span to guide an agent's container build changes
  • Conduct a mobile code review and submit structured comments directly from a phone
  • Collaborate with a remote agent on files hosted over SSH by forwarding the editor through an HTTP tunnel

When to use it

  • When an agent workflow requires precise, location-anchored human feedback rather than chat-level instructions
  • When working with long files where collapsing and navigating sections improves review efficiency
  • When conducting reviews on a mobile device or from a remote machine via SSH port forwarding
  • When the target file is source code, a config format, or documentation that benefits from syntax highlighting

When not to use it

  • When the agent workflow has no human-in-the-loop review step and fully automated processing is expected
  • When feedback needs to be delivered programmatically without opening a browser editor
  • When the review target is a binary file or format not supported by the text editor
  • When no tools are listed, so agent-side tool invocation is not available through this skill