Project Alignment

Align is a project-wide health check that adapts to where your project is in its lifecycle:

Pre-PRD (phases planning / imported) — evaluates your roadmap for consistency, completeness, and clarity before generating the first PRD
Post-PRD (phases structured / executing) — scans the project as a whole to surface drift, scope gaps, dependency issues, and technical-debt hotspots

Trinity picks the mode automatically based on the project phase — you don't toggle it.

Accessing Align

Click Align in the sidebar's Project section. It unlocks at the planning phase.

How the Scan Works

Align runs an agent that pulls context through the Trinity MCP server:

Always — your vault index (project knowledge + gotchas filtered to your stack)
Post-PRD only — full PRD structure + stories, recent agent handoffs, daily + PRD recaps
Pre-PRD only — all 5 roadmap sections (vision, phases, milestones, architecture, design)
Specific mode — vault search by keyword + targeted handoff history when relevant

The scan takes a minute or two. Progress is shown in the sidebar task indicator — you can navigate away safely.

Running a Scan

Sweep Mode

Leave the chat blank and click the sweep button:

Pre-PRD — "Full Alignment"
Post-PRD — "General Check"

Trinity does an end-to-end sweep with no guidance beyond the current project state.

Type a concrete concern into the chat (e.g., "The auth flow seems over-engineered", "We're behind on the core features", "API stories feel disconnected from the auth phase"). Trinity focuses analysis on that area while still checking for related issues. You can go back and forth in the chat to sharpen the concern before running the sweep.

Understanding Results

Health Score (0–100)

A single overall number, color-coded:

Score	Color
85+	green
70–84	yellow
50–69	orange
<50	red

The score is a quick reference — the real value is in the individual findings.

Findings

Each finding includes:

Title — short summary of what was found
Severity — how important it is to address
Category — area it relates to (architecture, security, performance, scope, etc.)
Description — detailed explanation of the issue
Recommendation — suggested approach to fix it

Summary

A high-level narrative summarizing the project's state and the most important next actions.

Acting on Findings

Each finding has a checkbox. Select the ones you want to act on, then use the mode-specific button at the bottom:

Pre-PRD → "Refine Roadmap" — jumps back to the onboarding roadmap with the selected findings highlighted so you can address them before generating a PRD
Post-PRD → "Implement" — sends the selected findings to Architect as seed context. Architect classifies scope, proposes stories, and shows a preview before saving. This is the primary path from "Align found a gap" to "stories to fix it".

You can also use findings for manual review without acting on them — they're useful for prioritizing technical debt and understanding what needs attention.

Chat Follow-Up

After the initial scan, the chat stays open so you can dig deeper:

"Tell me more about the authentication finding"
"What's the impact of not fixing the deprecated dependency?"
"How would you prioritize these findings?"

The chat retains the full scan context, so follow-ups are grounded in the same analysis.

Before a new PRD — understand the current state before Architect proposes new work
After a sizable chunk of execution — verify nothing has drifted from the original intent
Periodically — catch compounding issues early
When something feels off — run a specific-concern check to name it

Align vs. Audit vs. Auditor

Three different checks, one per layer of the system:

Feature	What it checks	Where it runs
Align	Roadmap + project health (scope, drift, deps, planning)	Sidebar — on demand
Audit	Codebase quality (architecture, security, patterns)	Sidebar — on demand; see Audit Tracking
Auditor	Per-story code review	Inside the story pipeline — automatic

They complement each other: Align catches planning and scope issues, Audit catches codebase-level quality gaps, and the Auditor catches story-level problems as they happen.