agentsclimarketplace
Sign in

Brief

Skill lwalden/AIAgentMinder/skills/brief

Product brief and roadmap creationFrom its SKILL.md

Install
npx -y skills add lwalden/AIAgentMinder --skill brief

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

11.2 KB, as published. Nobody here has run it

/aiagentminder:brief - Product Brief & Roadmap Creation

Create or update docs/strategy-roadmap.md — the project's north star, the "why" behind every development decision.

Before Starting

Read CLAUDE.md to understand the project identity (name, type, stack). If Project Identity is still placeholder brackets, run /aiagentminder:setup first.

Note: Claude Code has a built-in /plan command that toggles Plan Mode (read-only exploration). This command (/aiagentminder:brief) is different — it produces a product brief and strategy roadmap.

Step 0: Assess Starting Point

Ask the user where they are:

A) Rough concept -- "I have a vague idea but need help figuring everything out" B) Clear idea, no details -- "I know what I want but haven't worked out specifics" C) Partial plan or spec -- "I have some docs/notes -- help me fill in gaps" D) Detailed plan or spec -- "I have a writeup -- translate it into a roadmap" E) Existing project -- "The project already has code -- I'm adding AIAgentMinder for governance"

Starting PointApproach
A) Rough conceptFull interview (all rounds) plus exploratory questions
B) Clear ideaFull interview as written below
C) Partial planRead shared docs, ask only about gaps
D) Detailed planGenerate roadmap directly, clarify ambiguities only
E) Existing projectCodebase audit + current-state interview; skip product brief, generate filled state files

Starting Point E: Existing Project

Skip the product planning interview. The goal is to capture current state, not design a product.

Step E1: Audit the Codebase

Read key files to understand what exists:

  • CLAUDE.md (if present) -- existing project identity or instructions
  • package.json, pyproject.toml, go.mod, Cargo.toml, or equivalent -- stack and dependencies
  • README.md -- project description
  • Directory structure -- infer architecture

Summarize what you found: language, framework, rough architecture, apparent project stage.

Step E2: Interview (one round)

Ask in a single grouped message:

  • What phase is this project in? (early prototype / active development / maintenance)
  • What's working and stable?
  • What's currently in progress or incomplete?
  • What are the immediate next priorities?
  • What significant decisions have already been made? (stack choices, auth approach, DB, APIs, key libraries) -- these become DECISIONS.md entries
  • Any known blockers or open questions?
  • Any governance features you want to disable? (TDD, sprint planning, architecture fitness rules are all enabled by default)

Step E3: Generate State Files

Do NOT generate docs/strategy-roadmap.md unless the user asks. Instead:

  1. Capture current state in docs/strategy-roadmap.md reflecting actual project status:

    • Phase based on what you learned
    • Active Tasks = what's in progress right now
    • Current State = what's working / partial / broken
    • Next Priorities = the immediate next steps the user described

  2. Seed DECISIONS.md with retroactive ADR entries for each significant decision identified in Step E2. Use the project's format (Lightweight or Formal -- ask if unknown). Each entry needs alternatives considered and tradeoffs, even if reconstructed from context.

  3. Populate CLAUDE.md Project Identity with actual values from the audit.

  4. Set up SPRINT.md (if not already present — setup may have created it):

    • Copy SPRINT.md from ${CLAUDE_PLUGIN_ROOT}/templates/SPRINT.md if SPRINT.md does not already exist in the project root.
    • Add @SPRINT.md to CLAUDE.md after the Context Budget table (Claude Code's native import syntax — loads SPRINT.md every session when the file exists).
    • Add SPRINT.md row to CLAUDE.md Context Budget table: | SPRINT.md | ~35 lines active | Archived when sprint completes |.
    • Add reminder to Human Actions: "Review and approve sprint specs before Claude begins coding — every sprint starts with your approval."

    Do NOT install code-quality.md, sprint-workflow.md, or architecture-fitness.md into .claude/rules/. These rules are embedded inline in the plugin's agents/dev.md and agents/sprint-master.md. They load automatically when the user runs claude --agent dev or claude --agent sprint-master. Installing copies into .claude/rules/ creates stale duplicates that drift from the agent definitions.

  5. Ask: "Do you want a docs/strategy-roadmap.md too? It's optional for existing projects -- useful if you want a north-star doc for future phases."

  6. Tell the user: "AIAgentMinder is set up. Run /aiagentminder:handoff at the end of each session to checkpoint decisions and key context."

Question Flow

Ask questions in grouped rounds, not one at a time. Adapt based on project type.

Pre-Round (Starting Point A only): Explore the Idea

  • Are there existing tools or products that solve a similar problem? What's wrong with them?
  • What would make someone choose your tool over doing nothing?
  • Is this a "scratching your own itch" project or aimed at others?

Round 1: Core Understanding

  • What does this project do? (elevator pitch)
  • Who will use it?
  • What's the core problem it solves?
  • What makes it different from alternatives?

Round 2: Scope & Technical

  • 3-5 must-have features for v1?
  • Features that can wait?
  • What is explicitly OUT of scope? (things this project will never do)
  • Constraints? (budget, hosting, compliance, timeline)
  • External services or APIs needed?
  • MCP servers? (database tools, browser automation, etc. -- or "none")
  • Target launch date?

Round 3: Governance Setup

All governance features are enabled by default via the plugin's agent definitions. Do not ask the user to opt into individual features.

Tell the user in one line:

"Governance features (TDD, sprint workflow, architecture fitness, scope guardian) load automatically via the plugin agents — run claude --agent dev for feature work, claude --agent sprint-master for sprints."

Install SPRINT.md (if not already present — setup may have created it):

  1. Check whether SPRINT.md already exists in the project root. If it does, skip to step 4.
  2. Copy SPRINT.md from ${CLAUDE_PLUGIN_ROOT}/templates/SPRINT.md.
  3. Add @SPRINT.md to CLAUDE.md after the Context Budget table (Claude Code's native import syntax — loads SPRINT.md every session when the file exists).
  4. Add to CLAUDE.md Context Budget table: | SPRINT.md | ~35 lines active | Archived when sprint completes |.
  5. Add to docs/strategy-roadmap.md Human Actions Needed: "Review and approve sprint specs before Claude begins coding — every sprint starts with your approval."

Do NOT install code-quality.md, sprint-workflow.md, or architecture-fitness.md into .claude/rules/. These are embedded inline in agents/dev.md and agents/sprint-master.md and load when those agents are invoked. Installing copies into .claude/rules/ creates stale duplicates that drift from the agent definitions.

Decision Forcing: Surface Hard-to-Reverse Choices

After Round 2 (or after gathering technical stack information in any starting point), identify decisions that are hard to reverse and surface them proactively. Do not wait for the user to raise them — the value is catching them before they're locked in.

For each significant technical choice, check whether it carries downstream consequences the user may not have considered. Raise it in a single grouped message:

Patterns that trigger a decision point:

ChoiceConsequence to surface
Relational DB (Postgres, MySQL)Search strategy — full-text search needs GIN index or external search service
JWT / stateless authNo server-side revocation — stolen tokens valid until expiry
NoSQL (MongoDB, DynamoDB)Joins are expensive — how will you handle relational data?
MonorepoCI complexity, shared dependency management overhead
MicroservicesNetwork latency, distributed tracing, eventual consistency
External auth provider (Auth0, Clerk)Vendor lock-in and per-user pricing at scale
ORM vs raw SQLORM hides query cost; raw SQL is harder to maintain
Server-side renderingSEO benefit, but state management becomes more complex
Third-party API as core dependencyOutage = your outage; rate limits = your limits

For each decision point identified, say:

"You've chosen [X]. This means [consequence]. [Optional: alternative and its tradeoff.] Do you want to decide [downstream question] now, or defer it?"

If the user decides now: log it. If they defer: add a <!-- TODO: --> marker to the roadmap.

Reversal cost language: When a decision is hard to reverse, say so explicitly. "This choice is hard to change after you have data in it" or "Swapping this later would require a database migration" makes the stakes concrete.

Round 4: Unknowns (only if gaps exist)

  • What decisions are you unsure about?
  • What needs research first?

After gathering answers, summarize and confirm: "Does this capture it? Anything to add?"

Document Generation

Fill in docs/strategy-roadmap.md using the lean template. Keep it brief — a product brief, not an enterprise strategy doc. The user can always expand sections later.

For Each MVP Feature, Include Acceptance Criteria

1. [Feature] -- Acceptance: [testable criterion]

For Non-Goals, Be Explicit

The "Out of Scope" section in the roadmap should list concrete things the project will NOT do, not vague disclaimers. "Won't support offline mode" is good. "Won't do everything" is useless.

For Unknowns, Use TODO Markers

<!-- TODO: [What needs deciding] | WHEN: [deadline] | BLOCKS: [what] -->

After Generation

  1. Write the completed docs/strategy-roadmap.md
  2. Ask: "Do you prefer lightweight one-liner ADRs or formal ADRs (Context / Decision / Consequences)?" Record the answer in DECISIONS.md as Format: Lightweight or Format: Formal
  3. Log decision points to DECISIONS.md — For each decision surfaced during the Decision Forcing step that the user resolved:
    • Record the decision with: what was chosen, alternatives considered, reversal cost, and why this choice was made.
    • Use this format (lightweight): 
      [TOPIC]: [Choice made]. Alternatives: [what else was considered]. Reversal cost: [high/medium/low — and why]. Rationale: [reason].
    • For deferred decisions: add a <!-- TODO: [decision question] | WHEN: [trigger] | BLOCKS: [what] --> marker in the roadmap's Open Questions section.
  4. Populate ## MVP Goals in CLAUDE.md with Phase 1 deliverables (3-5 testable bullet points)
  5. If MCP servers were mentioned, add **MCP Servers:** line to Project Identity in CLAUDE.md
  6. Summarize what was installed: "All governance features enabled — TDD, sprint planning, self-review, and architecture fitness rules. Governance rules live in .claude/rules/ and are loaded automatically each session. SPRINT.md is loaded via @import in CLAUDE.md."
  7. Tell the user: "Your roadmap is ready. When you're ready, say 'start a sprint' or 'begin Phase 1' and I'll propose issues with detailed specs for your review."

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.