Skip to content

feat(docs): reposition CoreRoom around issue-to-PR evidence #403

@spytensor

Description

@spytensor

Problem

CoreRoom's README and first public explanation describe many correct mechanisms — roles, TUI, journals, permissions, gates — but an external user does not immediately see the differentiated product value: CoreRoom is the AI engineering control plane that turns agent work into issue-scoped, PR-backed, evidence-carrying delivery.

This matters because Claude Code, Codex, Gemini CLI, and GitHub Copilot already cover "agent writes code" and increasingly cover multi-agent/background work. CoreRoom must make the evidence/control-plane delta obvious in the first screen.

Proposed solution

Reposition the README top section and first-run narrative around the killer loop:

issue / user-scoped intent -> @host intake -> scoped work -> agent execution -> PR evidence -> reviewer decision

Concretely:

  • Lead with the issue-to-PR evidence loop, not the roster/TUI feature list.
  • Add a short "Why not just Claude Code / Codex / Copilot?" contrast focused on governance, evidence, and cross-engine control.
  • Add a 90-second demo script showing a realistic issue, acceptance criteria, branch/PR evidence, risks, rollback, and tracker handling.
  • Keep install/update instructions accurate and avoid unverified market claims.

Alternatives considered

  • Keep the README as a feature inventory. Rejected: it makes CoreRoom look like a multi-agent terminal cockpit, which is not defensible against vendor CLIs.
  • Build more UI first. Rejected: users must understand why the product exists before more UI polish matters.

Scope target

v0.x or later (deferred is fine)

Acceptance criteria

  • README top section explains CoreRoom as an AI engineering control plane / evidence layer before listing role/TUI mechanics.
  • README includes a concise killer-loop walkthrough from issue or user-scoped intent to PR evidence.
  • README explicitly states what CoreRoom does not replace: agent runtimes, GitHub PRs, CI, reviewer judgment.
  • Any comparisons to Claude Code, Codex, Gemini CLI, or Copilot are phrased as positioning, not unverifiable superiority claims.
  • Install/update instructions and current version facts remain aligned with Cargo.toml, npm/package.json, and AGENTS.md.

Validation

  • python3 scripts/check-project-truth.py
  • git diff --check
  • Manual README pass: a new user can explain the killer loop in one sentence after reading the first screen.

Out of scope

  • No runtime behavior changes.
  • No new screenshots unless existing README visuals become misleading.
  • No claims of external adoption unless independently evidenced.

Metadata

Metadata

Assignees

No one assigned

    Labels

    area:evidenceEvidence packets, tracker updates, PR proofarea:githubGitHub Issues, PRs, tracker, CI integrationarea:hostHost-led intake, delegation, confirmation, and orchestrationcodex-readyCodex may pick up autonomouslydocumentationImprovements or additions to documentationpriority:highTruth-in-advertising or locked-spec gapsize:SLess than 1 daystatus:readyReady to be picked up

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions