cross-agent-plan
Get an independent implementation plan from a DIFFERENT AI agent over the repowire mesh before you build (e.g. have Pi or Codex draft an approach for Claude to…
Integration test for repowire peer-to-peer messaging. Supports claude-code, opencode, or mixed-agent-type testing with circle boundaries and cross-agent-type communication. Can run all modes in parallel via agent teams.
$ npx -y skills add prassanna-ravishankar/repowire --skill integration-test --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/integration-testContext preview
The summary Claude sees to decide when to auto-load this skill.
Integration test for repowire peer-to-peer messaging. Supports claude-code, opencode, or mixed-agent-type testing with circle boundaries and cross-agent-type communication. Can run all modes in parallel via agent teams.
name: integration-test description: Integration test for repowire peer-to-peer messaging. Supports claude-code, opencode, or mixed-agent-type testing with circle boundaries and cross-agent-type communication. Can run all modes in parallel via agent teams.
Unified integration test for peer-to-peer communication across agent types.
| Argument | Behavior | |----------|----------| | `claude-code` | Test Claude Code agent type only | | `opencode` | Test OpenCode agent type only | | `mixed` | Test cross-agent-type communication | | `all` or no argument | Run all three modes in parallel using agent teams |
When `all` is specified (or no argument), the lead agent should:
1. Run Phase 1 (environment discovery) and Phase 3 (pre-test teardown) itself 2. Ask user for project directories and opencode model (Phase 2) 3. Ensure daemon is running 4. Create tasks for each mode and spawn background agents in parallel:
5. Collect results and produce unified report
Use the Task tool with `run_in_background: true` to run modes in parallel:
Task(subagent_type="general-purpose", run_in_background=true, prompt="Run the claude-code integration test. [paste mode section]") Task(subagent_type="general-purpose", run_in_background=true, prompt="Run the opencode integration test. [paste mode section]")
Each agent gets the full context for its mode and runs independently.
Run these commands to understand current state. Keep all phases in the same shell so `REPOWIRE_BIN` continues to identify the exact binary under test.
# Pin every command in this workflow to one binary. Phase 4 replaces this with
# the freshly built checkout binary when a fresh install is requested.
REPOWIRE_BIN="$(command -v repowire || true)"
export REPOWIRE_BIN
# System checks
tmux list-sessions 2>/dev/null || echo "No tmux sessions"
which opencode 2>/dev/null || echo "opencode not in PATH"
which claude 2>/dev/null || echo "claude not in PATH"
# Daemon and peer status
curl -s http://127.0.0.1:8377/health 2>/dev/null | jq . || echo "Daemon not running"
curl -s http://127.0.0.1:8377/peers 2>/dev/null | jq '.peers[] | {name, status, circle}' || echo "No peers"
# Installation status
test -n "$REPOWIRE_BIN" && "$REPOWIRE_BIN" --version || echo "Repowire not installed"
jq -r '"Hooks: " + ((.hooks // {}) | keys | join(", "))' "$HOME/.claude/settings.json" 2>/dev/null || echo "No Claude hooks"
ls ~/.opencode/plugin/repowire.ts 2>/dev/null && echo "OpenCode plugin installed" || echo "No OpenCode plugin"Ask user for: 1. **Project directories** (2-3 depending on mode, or ask for all if running `all`) 2. **OpenCode model** (e.g., `anthropic/claude-sonnet-4-5-20250929`) — needed for opencode/mixed modes 3. **Fresh install?** — removes existing hooks/plugins and reinstalls
tmux kill-session -t circle-a 2>/dev/null || true
tmux kill-session -t circle-b 2>/dev/null || true
tmux kill-session -t opencode-test 2>/dev/null || true
tmux kill-session -t mixed-test 2>/dev/null || true
curl -s http://127.0.0.1:8377/health >/dev/null 2>&1 && {
"$REPOWIRE_BIN" peer prune --force 2>/dev/null || true
}
rm -f ~/.repowire/pending/*.json 2>/dev/null || true
rm -f ~/.repowire/sessions.json 2>/dev/null || true**IMPORTANT:** `repowire setup` calls `claude mcp add` which fails inside a Claude Code session. Run these commands in a **tmux pane**, not from within Claude Code.
# Run in a tmux pane (e.g. tmux send-keys to a spare window):
# Build from local source
cd /path/to/repowire
mkdir -p bin
(cd daemon-go && go build -o ../bin/repowire .)
REPOWIRE_BIN="$(pwd)/bin/repowire"
export REPOWIRE_BIN
# Setup hooks and MCP server
"$REPOWIRE_BIN" setup --no-service
# Verify
"$REPOWIRE_BIN" --version
jq -r '"Hooks: " + ((.hooks // {}) | keys | join(", "))' "$HOME/.claude/settings.json"
ls ~/.opencode/plugin/repowire.ts 2>/dev/null && echo "OpenCode plugin: OK"test -x "$REPOWIRE_BIN" || { echo "REPOWIRE_BIN is not executable: $REPOWIRE_BIN" >&2; exit 1; }
curl -s http://127.0.0.1:8377/health || {
nohup "$REPOWIRE_BIN" serve > /tmp/repowire-daemon.log 2>&1 &
sleep 3
}
curl -s http://127.0.0.1:8377/health | jq .---
**Requires:** tmux, claude CLI, Claude hooks installed **Tmux sessions used:** `circle-a`, `circle-b` **Project directories needed:** 3 (2 for circle-a, 1 for circle-b)
tmux new-session -d -s circle-a -n peer-a1 tmux new-window -t circle-a -n peer-a2 tmux new-session -d -s circle-b -n peer-b1 tmux send-keys -t circle-a:peer-a1 "cd $PROJECT_A1 && claude --dangerously-skip-permissions" Enter sleep 3 tmux send-keys -t circle-a:peer-a2 "cd $PROJECT_A2 && claude --dangerously-skip-permissions" Enter sleep 3 tmux send-keys -t circle-b:peer-b1 "cd $PROJECT_B1 && claude --dangerously-skip-permissions" Enter sleep 30 # Wait for sessions to initialize and hooks to register
# All 3 peers should be online with correct circles and peer_id format
curl -fsS http://127.0.0.1:8377/peers | jq -e --arg a1 "$PEER_A1" --arg a2 "$PEER_A2" --arg b1 "$PEER_B1" '
def valid($name; $circle):
[.peers[] | select(
.name == $name and
.status == "online" and
.circle == $circle and
(.peer_id | test("^repow-" + $circle + "-[a-f0-9]{8}$"))
)] | length == 1;
if valid($a1; "circle-a") and valid($a2; "circle-a") and valid($b1; "circle-b") then
[.peers[] | select(.name == $a1 or .name == $a2 or .name == $b1) |
{name, sMay the agents talk . Connect Claude Code, Opencode, Codex, Pi across projects, across machines and with your telegram
Repo: prassanna-ravishankar/repowire
Get an independent implementation plan from a DIFFERENT AI agent over the repowire mesh before you build (e.g. have Pi or Codex draft an approach for Claude to…
Get a code/PR/plan review from a DIFFERENT AI agent over the repowire mesh (e.g. have Codex review Claude's work). Use when you want an independent second…
Hand a task to another AI agent peer over the repowire mesh — reuse an existing peer on the chosen backend, or spawn one if none exists. Use when you want to…
Install or update repowire and its skill pack from inside an agent session — ensure the repowire CLI/hooks are set up, install the skills, and verify the mesh…
Reference for how to use the repowire mesh — ask/ack vs notify, broadcast, peer discovery, spawning, and cross-agent workflows. Use when you need to coordinate…