brief

Harness Brief — Create Structured Briefs

Synthesize a shaping conversation, raw idea, or exploration into a structured brief saved to the project's target directory.

Prerequisites

Check for .harness/kb/stack-profile.md. If missing, tell the user to run /harness:init first.

Context Loading

• Load ALL KB files from .harness/kb/ for comprehensive context
• Check documents/shapings/ (or the skillflow's target_dir) for existing briefs on related topics
• If there's an active skillflow, read prior step outputs

Process

Step 1: Find the Input

Determine where the brief content comes from:

1. Active conversation — a shaping discussion just happened (most common)
2. User provides description — "Write a brief about X"
3. Existing file — "Turn this spec/doc into a brief"

If no clear input, ask: "What idea or feature should this brief cover?"

Step 2: Load KB for Grounding

Based on the topic, load relevant KB files and search the codebase to ground the brief in reality:

• What data entities relate to this idea?
• What APIs or jobs already exist for this domain?
• What UI components could be reused?

Step 3: Draft the Brief

Structure:

1. Problem / Opportunity — What this solves and for whom
2. Key Findings — What was discovered, with citations to specific files, tables, components
3. What Exists Today — Concrete list of existing infrastructure that relates
4. What Would Need Building — New data models, APIs, UI, jobs, integrations
5. Feasibility Signal — Confidence (high/medium/low), effort (T-shirt size), impact, appetite ratings
6. Open Questions — Unresolved items that need answers
7. Recommended Next Step — Which skill to invoke next (/harness:ground, /harness:design, etc.)

Step 4: Review with User

Present the draft brief. Iterate 1-2 rounds based on feedback:

• Missing context?
• Scope too wide or too narrow?
• Effort estimate feel right?

Step 5: Save

Save to {target_dir}/brief-{slug}-{date}.md with frontmatter:

---
title: "{brief title}"
type: brief
date: {ISO date}
tags: [{relevant}, {tags}]
next-skill: ground
---

Also store via MCP: call harness_add_skillflow_step with skill_type: "brief", the user input, and the brief content as ai_output.

Guardrails

• Every "What Exists" claim must cite a specific file, table, or component
• Effort estimates should be T-shirt sized (S/M/L/XL), not time-based
• Don't propose solutions in the brief — that's what /harness:ground and /harness:design are for
• Briefs should be readable by non-technical stakeholders