Twelve skills.
One harness.

A Claude Code template. Idea → researched → shipped → remembered, in one session or many. The next session starts where this one ended.

$curl -fsSL https://raw.githubusercontent.com/Reef123/OpenVera/main/install.sh | bash
MIT · Claude Code 2.x · Read the docs →
§ 01 · The loop

Four skills, one round trip.

A vague idea becomes a shipped V0 in a session that remembers what just happened. Next session opens with that memory loaded.

/start-here Turn a half-formed idea into something to build. /research Multi-model research, saved as a doc. /build new From plan to working app, in one session. /doc-sync Save the session: state, log, roadmap. Next session opens with state loaded. STARTS AHEAD
Fig. 1. The loop. One session for V0, many for /build full.
§ 02 · The shift

Sessions that build on the last.

Claude Code resets every session. OpenVera keeps a structured memory that survives across them.

Every session starts from zero.

Without a harness, each run re-derives context, re-explains the codebase, and re-learns your conventions. Good decisions from yesterday are invisible today.

Every session builds on the last.

With Vera, commands, skills, and patterns persist. Prior research artifacts are citable. Decisions carry forward. The harness gets smarter with use.

WITHOUT OPENVERA S1 S2 S3 S4 S5 S6 S7 reset reset reset reset reset reset WITH OPENVERA S1 S2 S3 S4 S5 S6 S7 carry carry carry carry carry carry
Fig. 2. Session-to-session continuity. Isolated sessions reset; harnessed sessions accumulate state.
§ 03 · Boot sequence

Three tiers of memory.

What’s loaded up front, what’s fetched on demand, and what stays out of context until searched.

Core EVERY SESSION · ~300 LINES state.md · memory/patterns.md · relationships/user.md Recall ON DEMAND · LAZY LOAD ROADMAP.md · who-i-am/* · recent conversations Archival SEARCH ONLY · NEVER PRE-LOADED conversations/ · research papers · superseded plans
Fig. 3. Tiered context budget. Core is ambient; Recall is lazy; Archival is retrieval-only.
Core
Loaded every session
  • state.md
  • memory/patterns.md
  • relationships/user.md
Recall
Loaded on demand
  • ROADMAP.md
  • who-i-am/identity.md
  • who-i-am/voice.md
  • recent conversations
Archival
Search only, never pre-loaded
  • conversations/
  • research papers
  • superseded plans
§ 04 · Three primitives

Agents, Commands, Skills.

The harness is composed of three kinds of file. Each has a clear role; together they orchestrate work.

PrimitiveWhatWhereExample
Agents Autonomous actors in isolated contexts. Fresh memory, scoped tools, can be spawned in parallel. .claude/agents/<name>.md Research subagent, code reviewer
Commands Prompt templates injected into current context. Orchestrate workflows, trigger skills. .claude/commands/<name>.md /doc-sync, /commit
Skills Reusable knowledge packages. Auto-discoverable, invoked via /slash-command. .claude/skills/<name>/SKILL.md /research, /build, /improve
Orchestration: Command → Agent → Skill.
§ 05 · Skills

Twelve commands, one slash each.

Each skill is a knowledge package with a command surface. Invoke by typing its slash.

/start-here
Takes a vague idea and turns it into something buildable.
/scout
Quick research: Reddit, YouTube, web. 2–3 min.
/research
Deep multi-model research. Source registry, paper output.
/consult
Simulates a panel of domain experts, gives one recommendation.
/frame
Design system, architecture diagrams, wireframes.
/panel
Pressure-test the bet before /build. Two reviewers scan for blind spots.
/advisor
Checks a decision against project artifacts, reports mismatches.
/build new
V0: idea to working app in one session.
/build full
Full SDLC: PRD, spec, build, review. Multi-session.
/improve
Runs a skill, scores the output, proposes instruction fixes.
/curate
Weekly memory cleanup: prunes stale, merges duplicates.
/doc-sync
Updates state file, conversation log, roadmap.
§ 06 · Two build paths

Ship in a session, or plan across many.

Choose based on scope. Both converge on the same memory and pattern store.

/build new
One session, idea to working prototype.
For V0s and spikes. Collapses scoping, scaffolding, and shipping into a single run.
/build full
Multi-session, PRD to ship.
For production work. PRD, design, implementation, review, and release across persisted sessions.
/ BUILD NEW idea working prototype single session / BUILD FULL PRD design build review ship learn session 1 session 2 session 3 session 4 session 5
Fig. 4. Build journey. /build new collapses the arc; /build full persists it across runs.
§ 07 · Safety

What’s denied, what asks, what’s scoped.

Defaults come from .claude/settings.json and .claude/rules/. Edit per project as needed.

Deny

  • sudo
  • rm -rf · rm -fr
  • .env files
  • .pem · .key · .ssh
  • .aws credentials

Ask first

  • git reset --hard
  • git push -f (force variants)
  • git clean -f
  • kill · pkill · killall
  • .secrets files

Rules

  • File-Size Guard: caps on CLAUDE.md, state.md, patterns.md, ROADMAP.md.
  • Self-Audit: flags deny removals or ask downgrades.
  • Test Integrity: fix the code, not the assertions.

Untrusted content

  • Extract techniques, not conclusions.
  • Verify first, implement second.
  • External recommendations are input, not orders.
§ 08 · Install

One command.

$curl -fsSL https://raw.githubusercontent.com/Reef123/OpenVera/main/install.sh | bash
Prereqs: Claude Code CLI · Python 3.8+ · git

Optional API keys (OpenRouter, Google AI) unlock multi-model research and YouTube analysis.

COPIED