Log in

Shared Skills

Reusable workflows your team authors and runs together. Pre-connected to your stack, versioned, reviewed, owned.

A Shared Skill is the unit of repeatable team work in Headways. It packages the prompt, the context, the connected tools, the review gates, and the owner into one thing the whole team can run — not a private prompt sitting in one person's chat history.

Skills are how a team's best work patterns stop being someone's individual habit and start being something the whole team can use, edit, and improve.

What's in a Skill

Each Skill bundles everything the platform needs to run it end-to-end:

  • Instructions — The system prompt and behavior contract for the runtime
  • Context inputs — What the Skill reads from the Knowledge Graph and which Connector sources it can pull from
  • Drafting steps — What artifacts get produced (email reply, brief, deck section, CRM update, doc edit)
  • Review checkpoints — Where a human reviews, edits, or approves
  • Owner — Who's accountable for the outputs
  • Governance metadata — Version history, protection rules, approval chain
  • Quality signals — Acceptance rate, edit distance, downstream outcome, adoption by team

Lifecycle

  1. Draft — A teammate authors the Skill in the workspace, via the CLI, or by promoting a Shared Conversation that produced great work.
  2. Review — Submitted for review. Protection rules validate approval requirements before publication.
  3. Published — Available for the team or org to install. Visible in the Agent Workspace Skills tab.
  4. Active — Installed and producing outputs. Every run lands in the Output Ledger.
  5. Forked or Iterated — Teams fork, adapt, re-publish. Iteration is driven by ledger signals — acceptance rates, edits, discards — not by guess.
  6. Archived — Removed from the library when superseded. Run history is preserved for audit.

Forking and Reuse

Skills are designed to spread. A Q3 launch-brief Skill authored by one team can be forked by another, edited for their conventions, and re-published — with attribution intact. Forks carry their parent's lineage so the library stays honest about what came from where.

This is the mechanism that turns one team's working pattern into the org's.

How Skills Compose with the Rest of the Platform

  • Connectors define when a Skill runs — Slack mention, deal stage change, calendar event, inbound email.
  • Knowledge Graph provides the context every Skill draws from at runtime.
  • Agent Workspace is where the Skill's drafted output lands for review — Inbox for prepared work, Outbox for outbound drafts.
  • Shared Conversations carry the full thread, sources, and work product between teammates as a Skill executes.
  • Governance decides what publishes, what gates apply, who has to approve.
  • Output Ledger records every output, the review action, and downstream signals.
  • Daily Library Review uses ledger signals to promote winning Skills and retire the ones that aren't pulling weight.

Cross-Runtime

Skills are written once and run anywhere the team uses AI. The current first-class runtime is the Headways Agent Workspace. Additional runtime adapters — Claude Code, Cursor, Copilot, Codex — let the same Skill definition execute in an engineer's IDE without forking the source of truth. See Runtime Adapters.

Why "Shared" Matters

Many tools now offer a Skill library or prompt library. That isn't the product. What makes Headways different is everything around the Skill: it's shared, governed, measured, and improving. A Skill in isolation is a prompt. A Skill in Headways is a team asset.