Developer franciscoh017 has released Concision, a portable skill for AI agents designed to solve a persistent workplace communication problem: over-explaining technical work. The open-source tool, now available on GitHub under the MIT license, helps AI assistants adapt messages based on audience intent rather than just shortening text. It’s a clever hack for anyone tired of AI agents that sound like verbose documentation rather than competent colleagues.

The Problem With 'Just Be Brief'

The creator notes that receiving feedback to 'be more concise' is common, especially for engineers who deeply care about the nuances of their work. However, simply reducing word count often destroys clarity. A message to an engineer needs trade-off details, while a note to management needs a clear decision point. Concision addresses this by forcing the AI to first identify the recipient’s actual need before generating a response, ensuring the right details are included for the right stakeholder.

Frameworks and Real-World Testing

Built upon Harsha Perera’s 'The Art of Being Concise,' the skill uses four core elements: the real question, the audience, curiosity, and knowing where you stand. The developer demonstrated the skill’s utility using a fictional performance pilot where p95 search latency dropped from 820 ms to 510 ms. For engineering, the AI output focused on pending security checks for tenant isolation; for product, it framed the speed gain as a user experience improvement; for management, it highlighted that no production approval was requested yet. The facts remained identical, but the narrative shifted to fit the room.

Technical Implementation and Access

Concision is implemented as a SKILL.md file, making it compatible with various AI agent frameworks without requiring external APIs or connectors. Users can install it directly via the Skills CLI using the command 'npx skills add franciscoh017/concision'. The project is explicitly positioned as a work in progress, with the developer stating that AI should support human practice in communication, not replace the human’s responsibility for what is said. This aligns with the broader trend of 'agent skills' becoming modular, reusable components in the AI stack.

Key Takeaways

  • Concision is an open-source, portable skill for AI agents that adapts message tone and detail based on audience.
  • It moves beyond simple summarization by using a framework focused on the recipient's intent and decision needs.
  • Installation is frictionless via the Skills CLI, requiring no external APIs or complex connectors.
  • The tool is designed to augment human communication skills, not replace the user's judgment.

The Bottom Line

Most AI agents are just noisy mirrors of our own verbosity. Concision is a necessary patch for agents that need to stop lecturing and start reporting like professionals.