| name | alerts |
| description | Intelligent health dashboard with scoring, benchmarks, trends, and interactive
alert-by-alert workflow. Triggers: "alerts", "check alerts", "what needs
attention", "system health", "show warnings", "pending issues". Default mode
(--limited) checks 16 categories. Use --full for comprehensive reporting with
all 36 categories.
|
Alerts -- Intelligent Health Dashboard
When to Use
- User explicitly invokes
/alerts
When NOT to Use
- When a more specialized skill exists for the specific task
Overview
Intelligent health dashboard: computes health scores, rates against benchmarks,
shows trends via sparklines, groups related items, and uses interactive
alert-by-alert workflow.
Output: v2 JSON with {alerts:[], context:{}} per category, health scores,
benchmarks, trends, and session plans.
Usage
/alerts # Limited mode (default) - 16 categories
/alerts --full # Full mode - 36 categories
Workflow
Phase 1: Run & Parse
node .claude/skills/alerts/scripts/run-alerts.js --limited (or --full)- Parse v2 JSON from stdout
- Create session log:
.claude/tmp/alert-session-{YYYY-MM-DD-HHMM}.jsonl
- Load suppressions from
.claude/state/alert-suppressions.json
Phase 2: Dashboard Overview
Health: {grade} ({score}/100) | {errors} errors / {warnings} warnings / {info} info
Category scorecard table (Category | Score | Rating), then: "Found N alerts.
Walking through each one..."
Phase 3: Alert-by-Alert Loop
Sort: errors first, then warnings, then info. For each alert show context card
with severity badge, category, message, details, trend, suggested action.
AskUserQuestion options by severity:
- ERROR: Fix Now | Defer | Suppress (permanent)
- WARNING: Fix Now | Defer | Ignore (session) | Suppress
- INFO: Acknowledge | Ignore (session) | Suppress
Suppress requires reason. Fix Now executes immediately. Log every decision to
session JSONL.
Phase 4: Action Plan Summary
Show: Fixed N, Deferred N, Ignored N, Suppressed N. If deferred items: list
them, ask "Execute deferred fixes now?" (Execute all / selected / Skip).
Phase 5: Batch Execution
Execute deferred fixes in order, update session log with results.
Phase 6: System Self-Audit
Claude performs live analysis checking: checker coverage (no_data), suppression
health (stale >90d), score integrity (unmeasured weight), decision balance
(fix/defer/ignore/suppress ratio), trend health (declining scores), process gaps
(missing log files). Present with AskUserQuestion: Acknowledge / Create tasks /
Suppress.
Phase 7: Cleanup & Verification
- Write
.claude/alerts-acknowledged.json
- Save new suppressions
- Clear resolved warnings from
.claude/hook-warnings.json
- Offer re-run with delta display
Suppression System
File: .claude/state/alert-suppressions.json. Match by category + regex on
message. Expired suppressions skipped. Filtered after checkers run, before score
computation.
Modes
Limited (16 categories): Code Health, Security, Session Context, Debt,
Learning, Agent Compliance, Hook Warnings, Skip Abuse, Test Results, Hook
Health, Session State, Pattern Hotspots, Context Usage, Reviews Sync, Review
Archive Health, Cross-Document Dependencies.
Full (+20 categories): Debt Intake/Resolution, Doc Health, Roadmap, Review
Quality, Consolidation, Velocity, Session/Commit Activity, Roadmap
Validation/Hygiene, Trigger Compliance, Pattern Sync, Doc Placement, External
Links, Unused Deps, Review Churn, Backlog, GitHub Actions, SonarCloud.
Benchmarks
All ratings: Good / Average / Poor.
| Category | Metric | Good | Average | Poor |
|---|
| Debt | S0 items | 0 | -- | >0 |
| Debt | S1 items | <10 | -- | >10 |
| Debt | Resolution rate | >50% | >30% | <10% |
| Code | TS errors | 0 | <5 | >20 |
| Code | ESLint warnings | 0 | <10 | >50 |
| Tests | Pass rate | >98% | >90% | <80% |
| Security | Critical vulns | 0 | 0 | >0 |
| Security | High vulns | 0 | <2 | >5 |
| Learning | Effectiveness | >85% | >75% | <60% |
| Learning | Automation coverage | >40% | >25% | <10% |
| Velocity | Items/session | >5 | >2 | 0 |
| Reviews | Fix ratio | <15% | <25% | >35% |
| Reviews | Max rounds | <2 | <3 | >5 |
| Hook Health | Warnings (7d) | 0 | <5 | >15 |
| Hook Health | False positive % | 0% | <30% | >60% |
| Agents | Compliance % | 100% | >80% | <50% |
| Docs | Staleness (days) | <3 | <7 | >14 |
Health Score Weights
Core (70%): Code 15%, Security 15%, Debt 12%, Tests 10%, Learning 8%, Agents
4%, Session 3%, Hook Health 3%, Skip Abuse 2%.
New state (8%): Session State 3%, Pattern Hotspots 3%, Context Usage 2%.
Existing (9%): Velocity 3%, Reviews 3%, Docs 3%.
Full-mode only (13%): Debt Intake 2%, Roadmap 2%, GitHub Actions 2%,
SonarCloud 2%, plus 1% each for Trigger Compliance, Pattern Sync, Doc Placement,
External Links, Unused Deps, Review Churn, Backlog.
Scoring: Start at 100, deduct 30/error, 10/warning. Grades: A=90+,
B=80+, C=70+, D=60+, F=<60.
Scripts
node .claude/skills/alerts/scripts/run-alerts.js --limited
npm run alerts:cleanup
Exit code 1 if error-level alerts. Delta tracking: first run/day saves baseline
to .claude/state/alerts-baseline.json.
Integration
At session start: run /alerts (limited), walk through workflow, fix issues,
offer re-run to verify.
Version History
| Version | Date | Description |
|---|
| 1.1 | 2026-02-25 | Trim to <500 lines: condense benchmark tables and weights |
| 1.0 | 2026-02-25 | Initial implementation |
