Fleet Alert

Raise alerts when you notice problems the fleet or human should know about.

Severity Routing

Severity	Board Memory	IRC #alerts	IRC #fleet	Task Blocker
🔴 Critical	✅	✅	✅	✅ (if task-related)
🟠 High	✅	✅	—	Optional
🟡 Medium	✅	—	✅	—
🟢 Low	✅	—	—	—

Alert Categories

Category	Examples
security	CVE in dependency, exposed secret in code, auth bypass, insecure config
quality	No tests for critical module, <50% coverage, code smell, missing validation
architecture	Tight coupling, scaling concern, design debt, circular dependency
workflow	Broken automation, manual step that should be scripted, missing tooling
tooling	Missing skill, agent can't complete task due to tool gap, broken script

How to Alert

Step 1: Post to board memory

Use template agents/_template/markdown/memory-alert.md:

curl -s -X POST "$BASE_URL/api/v1/agent/boards/$BOARD_ID/memory" \
  -H "X-Agent-Token: $AUTH_TOKEN" -H "Content-Type: application/json" \
  -d '{
    "content": "## ⚠️ {severity}: {title}\n\n**Project:** {project}\n**Found by:** {agent_name}\n**Severity:** {severity}\n**Category:** {category}\n\n### Details\n{evidence with URLs}\n\n### Recommended Action\n{what should be done}\n\n### References\n- {urls}\n\n---\nTags: alert, {severity}, {category}, project:{project}",
    "tags": ["alert", "{severity}", "{category}", "project:{project}"],
    "source": "{agent_name}"
  }'

Step 2: Post to IRC (if severity warrants it)

# Critical → #alerts
bash scripts/notify-irc.sh --agent "{agent_name}" --event "CRITICAL" --title "{title}" --url "{reference_url}"

# High → #alerts
bash scripts/notify-irc.sh --agent "{agent_name}" --event "HIGH" --title "{title}" --url "{reference_url}"

# Medium → #fleet
bash scripts/notify-irc.sh --agent "{agent_name}" --event "MEDIUM" --title "{title}" --url "{reference_url}"

Step 3: If task-related and critical, add blocker

curl -s -X POST "$BASE_URL/api/v1/agent/boards/$BOARD_ID/tasks/$TASK_ID/comments" \
  -H "X-Agent-Token: $AUTH_TOKEN" -H "Content-Type: application/json" \
  -d '{"message": "## 🚫 Blocked by Alert\n\n**Alert:** {title}\n**Severity:** {severity}\n\n{details}\n\nThis task cannot safely proceed until the alert is resolved."}'

Step 4: Optionally create a remediation task

If the alert requires specific work, use fleet-task-create to propose a task.

What to Include in Alerts

Good alert:

## ⚠️ HIGH: NNRT dependency `pydantic` has known CVE-2024-XXXXX

**Project:** nnrt
**Severity:** high
**Category:** security

### Details
`pydantic==2.5.0` in pyproject.toml has CVE-2024-XXXXX (arbitrary code execution
via crafted JSON). See: https://nvd.nist.gov/vuln/detail/CVE-2024-XXXXX

### Recommended Action
Upgrade to `pydantic>=2.5.3` which patches the vulnerability.

### References
- [pyproject.toml](https://github.com/cyberpunk042/Narrative-to-Neutral-Report-Transformer/blob/main/pyproject.toml)
- [CVE-2024-XXXXX](https://nvd.nist.gov/vuln/detail/CVE-2024-XXXXX)

Bad alert:

Found a security issue in the project.

Rules

• Be specific — include evidence, file paths (as URLs), CVE IDs
• Be actionable — say what should be done, not just what's wrong
• Use the right severity — critical means "stop and fix NOW"
• Don't alert for known issues — check board memory first
• Include references — URLs to files, CVEs, docs, issues