Board of Governance — Smoke Test

Validate that all agents can read their identity, receive a brief, and produce a report. This is the canonical way to verify a board installation works.

Step 1: Read the Board

Read .board/board/BOARD.md for agent list, invocation commands, and board user. If BOARD.md does not exist, error: "No board found. Run /setup first."

Step 2: Pre-flight Checks

Before running agents, verify the environment is functional:

1. •Board user exists: id $BOARD_USER — error if missing
2. •Agent directories complete: Each agent has CLAUDE.md, inbox/, outbox/
3. •Outbox writable by board user: sudo -u $BOARD_USER test -w $AGENT_DIR/outbox
4. •CLIs on PATH for board user: sudo -u $BOARD_USER bash -c 'which claude' (or codex)
5. •Credentials exist: Claude: ~$BOARD_USER/.claude/.credentials.json. Codex: ~$BOARD_USER/.codex/auth.json

If any check fails, stop and report the specific failure with a fix suggestion.

Step 3: Prepare Agents

For each agent listed in BOARD.md:

1. •

   Write test brief to inbox/brief.md:

   Smoke test: confirm your identity. State your name, thinking style, and which CLI you are running on. Write your response to outbox/report.md.

2. •

   Write minimal context to inbox/context.md:

   This is a smoke test. No domain context needed.

3. •

   Clear stale reports: rm -f $AGENT_DIR/outbox/report.md

Step 4: Sequential Test

Run each agent one at a time using its invocation command from BOARD.md. Sequential execution isolates failures — if agent 1 fails, you know it's not a parallelism issue.

For each agent:

1. •Record start time
2. •Run the invocation command (with a shorter timeout — 120s instead of 900s)
3. •Record end time
4. •Check exit code
5. •Verify outbox/report.md exists and is non-empty ([ -s ... ])
6. •Read first 3 lines of the report
7. •Record: PASS or FAIL with timing

If any agent fails:

• •Report the failure with exit code and any error output
• •Point to /debug for troubleshooting
• •Continue testing remaining agents (don't stop at first failure)

Step 5: Parallel Test

Only if all agents passed sequential. If any failed, skip this step.

1. •Clear all outbox reports again
2. •Rewrite test briefs (same content)
3. •Run all agents simultaneously using the parallelism pattern from BOARD.md (shorter timeout — 120s)
4. •Wait for all to complete
5. •Verify all reports exist and are non-empty
6. •Record wall-clock time for the parallel run

Step 6: Report Results

Present a summary table:

## Smoke Test Results

| Agent | CLI | Sequential | Parallel | Time (seq/par) |
|-------|-----|-----------|----------|----------------|
| pragmatist | claude | PASS | PASS | 15s / 14s |
| systems-thinker | codex | PASS | PASS | 25s / 22s |
| skeptic | claude | PASS | PASS | 16s / 15s |

**Board status:** All agents operational ✓

If any failures, include error details and /debug references.

Step 7: Cleanup

Remove smoke test briefs and reports (they're not real review artifacts):

for agent in $BOARD/board/*/; do
  name=$(basename "$agent")
  grep -q "Smoke test" "$agent/inbox/brief.md" 2>/dev/null && rm -f "$agent/inbox/brief.md" "$agent/inbox/context.md" "$agent/outbox/report.md"
done