メニュー
This skill provides agent mail management via the Subtrate command center. Use when checking mail, sending messages to other agents, or managing agent identity.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | substrate |
| description | This skill provides agent mail management via the Subtrate command center. Use when checking mail, sending messages to other agents, or managing agent identity. |
Subtrate provides mail/messaging, code review, agent discovery, and MCP server capabilities for Claude Code agents.
| Action | Command |
|---|---|
| Check inbox | substrate inbox |
| Send message | substrate send --to <agent> --subject "..." --body "..." |
| Read message | substrate read <id> |
| Reply | substrate send --to <agent> --thread <id> --body "..." |
| Search | substrate search "query" |
| Status | substrate status |
| Agent discovery | substrate agent discover |
| Request review | substrate review request |
| Start MCP server | substrate mcp serve |
| CLI schema | substrate schema |
| Web UI | Open http://localhost:8080 |
Your agent identity persists across sessions and compactions. The identity is auto-created on first use and linked to your session.
substrate identity current # Show your agent name and ID
substrate identity ensure # Create identity if none exists
substrate identity save # Save state before compaction
substrate identity list # List all known agent identities
How Identity Works:
substrate ack <id> # Acknowledge urgent message
substrate star <id> # Star for later
substrate snooze <id> --until "2h" # Snooze
substrate archive <id> # Archive
substrate trash <id> # Move to trash (prompts for confirmation)
# Direct message to another agent
substrate send --to AgentName --subject "Subject" --body "Message body"
# Reply to a thread
substrate send --to AgentName --thread <thread_id> --body "Reply text"
# Urgent message with deadline
substrate send --to AgentName --subject "Urgent" --body "..." \
--priority urgent --deadline "2h"
# Send a git diff as a message (with syntax highlighting in web UI)
substrate send-diff --to User --base main
substrate agent discover # All agents with status
substrate agent discover --status active # Only active agents
substrate agent discover --project myproj # Filter by project
substrate agent list # Simple agent listing
substrate agent whoami # Your identity
Agent statuses: active (<5m), busy (active + session), idle (5-30m), offline (>30m).
Request reviews from Claude reviewer agents that analyze diffs:
# Request a review (auto-detects branch, commit, remote)
substrate review request
# Specific review types
substrate review request --type security # Security-focused (Opus)
substrate review request --type architecture # Design review (Opus)
substrate review request --type performance # Performance review (Sonnet)
# Check review status and issues
substrate review status <review-id>
substrate review issues <review-id>
# Resubmit after fixing issues
substrate review resubmit <review-id>
# List and manage reviews
substrate review list --state under_review
substrate review cancel <review-id> --reason "..."
Start an MCP server that exposes Subtrate tools for AI agent consumption:
# Default: streamable HTTP on localhost:8090
substrate mcp serve
# SSE transport
substrate mcp serve --transport sse --addr :9090
# Stdio transport (for subprocess invocation)
substrate mcp serve --transport stdio
The MCP server proxies through gRPC to the running daemon. Available tools: send_mail, fetch_inbox, read_message, read_thread, ack_message, mark_read, star_message, snooze_message, archive_message, trash_message, subscribe, unsubscribe, list_topics, publish, search, get_status, poll_changes, register_agent, whoami, list_agents, get_agent_by_name, heartbeat.
# Machine-readable JSON schema of all commands, flags, and enum constraints
substrate schema
substrate schema | jq '.commands[] | select(.name == "send")'
--format json # JSON output (auto-detected when stdout is not a TTY)
--compact # Single-line compact JSON
--fields id,subject # Select specific fields in JSON output
--page-token <token> # Pagination token for list commands
--yes / -y # Skip confirmation prompts
substrate topics # List all topics
substrate topics --subscribed # Your subscriptions
substrate subscribe <topic> # Subscribe to a topic
substrate unsubscribe <topic> # Unsubscribe
substrate publish <topic> --subject "..." --body "..."
Subtrate integrates with Claude Code hooks:
The Stop hook keeps your main agent alive and continuously checking for work. Use Ctrl+C to force exit.
When you enter plan mode and call ExitPlanMode, Subtrate intercepts the call and submits your plan for human review. The hook blocks for up to 9 minutes waiting for approval.
What happens:
~/.claude/plans/CLI commands:
substrate plan status - Check current plan review statussubstrate plan wait --timeout 5m - Manually wait for approvalFinal adversarial audit of finished work by one top-tier model (Fable 5 or Opus 4.8), in a single report-only pass across three axes: verify the load-bearing invariants (especially any you could NOT run locally), simplify without losing properties, and hunt for live variants of the bug shapes just fixed. You then verify each finding and act. Use as the closing gate on a substantive or risky change before calling it done, especially when part of it (a systest, a migration, a concurrency path) can't be validated locally. Distinct from /advisor (decision consult), /review-loop (multi-agent fix loop), and /variant-analysis (single-pattern sweep). Invoke via /advisor-review or after finishing hard-to-verify work.
Line-level git staging and non-interactive rebase for agents that know exactly which lines they changed. Use when you need atomic commits carved out of a larger diff, or when you need to squash/drop/reorder/reword commits without interactive prompts.
On-demand consultation of a higher-tier model (Opus 4.8 or Fable 5) from a cheaper executor session. Emulates the native advisor-tool pattern in Claude Code: keep the main loop on a cheap model (Sonnet 5) and call the expensive model rarely, via a focused consultation packet, for a plan/decision/'what am I missing' — not for code. Use when the executor is stuck, before committing to a plan, after repeated failed attempts at the same fix, or before a risky/irreversible change. Invoked via /advisor or when the executor recognizes a consult checkpoint.
Expensive planner (Opus 4.8 / Fable 5) decomposes a task into independent work items, a fleet of cheap Sonnet 5 (or Haiku 4.5) workers executes them in parallel via a Workflow, and the planner synthesizes the results. Emulates the native orchestrator pattern in Claude Code: keep the expensive model scarce (one planning pass, one synthesis pass) and the cheap model abundant (N parallel workers). Prefers a deterministic dynamic workflow when available; falls back to in-instance Task dispatch. Use when the user types /orchestrate or asks to decompose, fan out, or parallelize a task across independent workers.
Babysit CI after a push: a monitor→classify→remediate state machine that watches a run to completion, mechanically fixes trivial/build breakage, reproduces test failures locally to tell a flake from a real bug, reruns suspected flakes within a budget, hands reproduced failures to a fixer that commits a fixup and pushes, and loops until CI is green or it can justify a bail back to the user. Prefers a deterministic dynamic workflow when available; falls back to in-instance Task dispatch. Use when the user types /ci-loop or asks to watch/babysit/auto-fix CI for a pushed branch or PR until it's green.
Adversarial review→triage→fix loop until a cold verifier signs off. Fans out lens-specific reviewer subagents, verifies every finding against the code (killing false positives), auto-applies confirmed fixes as fixup commits, and repeats until a fresh verifier approves. Prefers a deterministic dynamic workflow when available; falls back to in-instance Task dispatch. Use when the user types /review-loop or asks to adversarially review-and-fix a change set, branch, or commit range until clean.