Awesome-omni-skill project-spec

Create a project spec for AI agents to work from autonomously.

install
source · Clone the upstream repo
git clone https://github.com/diegosouzapw/awesome-omni-skill
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/diegosouzapw/awesome-omni-skill "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data-ai/project-spec" ~/.claude/skills/diegosouzapw-awesome-omni-skill-project-spec && rm -rf "$T"
manifest: skills/data-ai/project-spec/SKILL.md
source content

Project Spec

What to build, how it fits together, how to keep going.

Process

  1. Explore: Check context—files, docs, commits, patterns.
  2. Ask: Ask many questions (multiple choice when possible) until you can fill every section.
  3. Approaches: If multiple valid paths exist, present 2-3 options. Lead with your recommendation.
  4. Derive: Map conversation to spec sections:
    • Mission: what, who, why
    • Architecture: technical decisions
    • Components: what exists
    • Constraints: non-negotiables
    • Priorities: current focus
    • Standards: quality bar
  5. Write: Output
    PLAN.md
    with spec and progress in one file.

Spec Structure (under 100 lines)

Instructions:

  • Read Progress section to know what's done, check git log for context
  • Build the next item from Priorities—do not ask what to work on
  • Make decisions from existing code patterns
  • Commit at natural breakpoints (test passing, working feature, integration complete)
  • Update Progress section after each commit, continue to next item
  • When priorities clear, surface new work
  • PLAN.md is a living document—update as work evolves
  • No clarifying questions—ship working features

Mission: 1-3 lines. What, who, why.

Architecture: Technical decisions stated, not debated.

Components: What exists. One line each—name and essence.

Constraints: Non-negotiables only.

Priorities: Current focus. Replenishes.

Standards: Quality bar. Improve over time.

Progress: What's done, what's current, what's next. Update after each commit.

Writing Principles

Concrete, not vague

  • "CLI tool for rotating AWS credentials" not "developer utility"
  • "Postgres, Redis, S3" not "appropriate storage solutions"

No filler

  • Cut: "Prefer working software over perfection", "Start simple", "Future integrations"