Agent Inbox

Check for messages from autonomous agents at session start and process completion notifications.

Quick Start

Most common usage:

# List all messages
ailang messages list

# Show only unread messages
ailang messages list --unread

# Read full message content
ailang messages read MSG_ID

# Acknowledge (mark as read)
ailang messages ack MSG_ID
ailang messages ack --all

# Send a message
ailang messages send user "Your message" --title "Title" --from "agent-name"

# Send with semantic envelope (v0.8.1+)
ailang messages send executor "Fix bug" --title "Bug" \
  --envelope-code internal/parser/parser.go \
  --envelope-context "reviewing ast type switches"

# Search by envelope space (v0.8.1+)
ailang messages search --space code "internal/types"
ailang messages search --space intent "fix crash"

# Triage unread messages by similarity (v0.8.1+)
ailang messages triage --cluster-by code

# Send bug/feature to GitHub (for cross-instance visibility)
ailang messages send user "Bug report" --type bug --github

Expected output (at session start):

📬 AGENT INBOX: 2 unread message(s) from autonomous agents
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

ID: msg_20251210_143021_abc123
From: sprint-executor
Title: Sprint M-S1 complete
Time: 2025-12-10T14:30:21Z

ID: msg_20251210_143055_def456
From: stapledon
Title: Parser Bug
Time: 2025-12-10T14:30:55Z
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

When to Use This Skill

Invoke this skill when:

• Session starts - First action in every Codex session (required by AGENTS.md)
• After handoffs - When you've sent work to autonomous agents
• Periodic checks - User asks "any updates from agents?"
• Debugging - To see agent communication history

Storage Backend — TWO stores, and the default is the private one

Public feedback, package feedback, coordinator completions and other machines' agent traffic all land in the canonical store. A bare ailang messages list reads local SQLite and will not show any of it — measured 2026-08-26, the session banner reported 16 unread while the canonical store held 74.

Any listing against a non-local store prints its store in the header:

  store: gcp (Firestore, project ailang-multivac, via AILANG_STORAGE_MESSAGING)

No store: line means you are reading local only. Two causes: the vars are not set, or the binary predates the plane switch (M-V1-SIMPLIFY-S3 M3) and ignores them silently. Control for the second — an invalid value must be REFUSED:

AILANG_STORAGE_MESSAGING=not-a-real-store ailang messages list --unread
# current binary: errors "invalid storage selection: AILANG_STORAGE_MESSAGING=..."
# old binary:     lists normally  -> your counts are local-only

• Accessible via: CLI (ailang messages) and Collaboration Hub dashboard
• Message statuses: unread, read, archived, deleted

Traps

• messages read <id> marks it read as a side effect — triaging by reading drains the queue. Read bodies from --json to inspect without acking.
• The list view truncates IDs to 8 chars, so every cloud message shows as (inbox_17). Take full IDs from --json. An ambiguous prefix errors loudly rather than acking the wrong message.
• ack --all spans every inbox by default, including other people's package feedback. Scope with --inbox, or ack by ID.

Available Commands

List Messages

ailang messages list                    # All messages
ailang messages list --unread           # Only unread
ailang messages list --inbox user       # Filter by inbox
ailang messages list --from agent-name  # Filter by sender
ailang messages list --json             # JSON output
ailang messages list --limit 50         # Limit results

Read Full Message

ailang messages read MSG_ID             # Full content, marks as read
ailang messages read MSG_ID --peek      # View without marking read
ailang messages read MSG_ID --json      # JSON output

Acknowledge Messages

ailang messages ack MSG_ID              # Mark specific message as read
ailang messages ack --all               # Mark all as read
ailang messages ack --all --inbox user  # Mark all in inbox as read

Un-acknowledge (Mark Unread)

ailang messages unack MSG_ID            # Move back to unread

Send Messages

# Basic send (local only - for coordination)
ailang messages send INBOX "message" --title "Title" --from "agent"

# With semantic envelope (v0.8.1+ - attaches embeddings for multi-space search)
ailang messages send INBOX "Fix bug" --title "Bug" \
  --envelope-code internal/parser/parser.go          # embeds code context
ailang messages send INBOX "Fix bug" --title "Bug" \
  --envelope-context "reviewing ast type switches"   # embeds session context

# With GitHub sync (for bugs/features - cross-instance visibility)
ailang messages send INBOX "message" --type bug --github
ailang messages send INBOX "message" --type feature --github
ailang messages send INBOX "message" --github --repo owner/repo

Import from GitHub

ailang messages import-github                    # Import from default repo
ailang messages import-github --repo owner/repo  # Specific repo
ailang messages import-github --labels bug,help  # Filter by labels
ailang messages import-github --dry-run          # Preview without importing

Workflow

1. Session Start Check (REQUIRED)

SessionStart hook runs automatically and shows unread messages.

If messages exist:

• Read and summarize each message to user
• Identify message type (completion, error, handoff)
• Ask user if they want action taken
• Acknowledge after handling: ailang messages ack --all

2. Process Completion Notifications

When agent reports completion:

# 1. Read the full message
ailang messages read MSG_ID

# 2. Review results mentioned in payload
ls -la eval_results/baselines/v0.4.2/

# 3. Report to user
echo "Sprint complete! Results at: eval_results/baselines/v0.4.2/"

# 4. Acknowledge after processing
ailang messages ack MSG_ID

3. Handle Error Reports

When agent reports errors:

# 1. Read the full error details
ailang messages read MSG_ID

# 2. Check logs if mentioned
cat .ailang/state/logs/sprint-executor.log

# 3. Diagnose and report to user
echo "Agent encountered error: Tests failing at milestone 3/5"

# 4. Either fix manually or send corrective instructions

4. Respond to Agent or User

# Send response to an agent
ailang messages send sprint-executor "Approved, proceed" \
  --title "Approval" --from "user"

# Send notification to user inbox
ailang messages send user "Issue resolved" \
  --title "Status update" --from "Codex"

GitHub Integration (Bi-directional)

Message Types and Routing

Routing guidance:

• Bugs and features → Use --github for visibility across all AILANG instances
• Coordination messages → Local only, for agent-to-agent communication
• Instructions from humans → Create GitHub issues, they'll be imported automatically

Sending to GitHub (Agent → GitHub)

# Bug reports and feature requests go to GitHub for visibility
ailang messages send user "Parser crash" --type bug --github
ailang messages send user "Need async support" --type feature --github

Importing from GitHub (GitHub → Local)

# Runs automatically on session start (if auto_import: true in config)
ailang messages import-github

# Or manually with filters
ailang messages import-github --labels help-wanted

Human Instructions via GitHub

You can write instructions as GitHub issues and have agents pick them up:

1. Create issue on GitHub with ailang-message label
2. Next session, import-github runs automatically
3. Issue appears in agent's inbox as a message
4. Agent reads and acts on the instructions

Configuration

Create ~/.ailang/config.yaml:

github:
  expected_user: YourGitHubUsername   # REQUIRED: Must match gh auth status
  default_repo: sunholo-data/ailang   # Default repo for issues
  create_labels:
    - ailang-message
  watch_labels:
    - ailang-message
  auto_import: true                   # Auto-import on session start

Prerequisites:

1. Install GitHub CLI: brew install gh
2. Authenticate: gh auth login
3. Check account: gh auth status
4. Switch if needed: gh auth switch --user USERNAME

Auto-label creation: Labels are automatically created if they don't exist:

• from:agent-name (purple) - who sent the message
• bug (red), feature (cyan), general (light blue)
• ailang-message (blue) - identifies AILANG messages

Correlation IDs

Messages support correlation IDs for tracking handoff chains:

{
  "message_id": "msg_20251210_103045_abc123",
  "correlation_id": "sprint_M-S1",
  "from_agent": "sprint-executor",
  "to_inbox": "user",
  "title": "Sprint complete",
  "payload": "All milestones complete"
}

Benefits:

• Track entire workflow: design-doc → sprint-plan → execution
• Filter messages by workflow
• Debug multi-agent interactions
• Resume work from where you left off

For complete specification, see resources/message_format.md

Message Types (Payloads)

Completion Notification

{
  "type": "sprint_complete",
  "correlation_id": "sprint_M-S1",
  "payload": {
    "sprint_id": "M-S1",
    "milestones_complete": 5,
    "result": "All tests passing"
  }
}

Error Report

{
  "type": "error",
  "correlation_id": "sprint_M-S1",
  "payload": {
    "error": "Tests failing: 5 benchmarks broken",
    "details": ".ailang/state/logs/sprint-executor.log"
  }
}

Handoff Instruction

{
  "type": "plan_ready",
  "correlation_id": "sprint_M-S1",
  "payload": {
    "sprint_id": "M-S1",
    "plan_path": "design_docs/planned/M-S1-plan.md"
  }
}

Resources

Message Format Reference

See resources/message_format.md for complete message format specification.

Troubleshooting Guide

See resources/troubleshooting.md for common issues and solutions.

CLI Command Reference

Aliases: msg is an alias for messages

ailang msg list        # Same as: ailang messages list

Package Coordination Messages (M-PKG-MSG)

Package-scoped messages use typed inbox addressing for multi-agent coordination:

Inbox addressing: pkg:vendor/name, workspace:name, team:name

Auto-emitted on publish: upgrade-available, interface-change-notice, effect-widening-warning

11 message kinds: upgrade-available, interface-change-notice, effect-widening-warning, compatibility-request, compatibility-report, contract-regression, migration-request, deprecation-notice, upgrade-complete, blocked, superseded

Lifecycle: open → acknowledged → in_progress → completed (or blocked/rejected/superseded)

Notes

• Required by AGENTS.md: Session start check is mandatory
• SQLite backend: All messages in ~/.ailang/state/collaboration.db
• Hook integration: SessionStart hook auto-imports GitHub issues and shows unread
• Auto-marking: Messages marked as read when using ailang messages read
• Message lifecycle: Unread → Read → Archived (optional)
• GitHub sync: Optional, for bugs/features that need cross-instance visibility