Agent Skills › AlphaGBM/skills › alphagbm-health-check

alphagbm-health-check

GitHub

对研究知识库进行定期审计，检测过期档案、论点偏离和孤立页面，生成0-100健康评分及具体修复建议。适用于知识库状态审查或手动触发报告。

skills/alphagbm-health-check/SKILL.md AlphaGBM/skills

Trigger Scenarios

health check my research what's stale any drift generate a report 研究库体检过期档案论据偏离健康体检 audit

Install

npx skills add AlphaGBM/skills --skill alphagbm-health-check -g -y

More Options

Use without installing

npx skills use AlphaGBM/skills@alphagbm-health-check

指定 Agent (Claude Code)

npx skills add AlphaGBM/skills --skill alphagbm-health-check -a claude-code -g -y

安装 repo 全部 skill

npx skills add AlphaGBM/skills --all -g -y

预览 repo 内 skill

npx skills add AlphaGBM/skills --list

SKILL.md

Frontmatter

{
    "name": "alphagbm-health-check",
    "description": "Weekly diagnostic report on a user's research knowledge base — flags stale profiles (not updated in weeks), thesis drift (AI detects original premise no longer holds), and orphan pages (profiles with no thesis, themes with missing profiles). Returns an overall 0-100 health score with specific action recommendations. Use when: auditing knowledge base state, triggering a fresh report, reviewing what needs attention. Triggers on: \"health check my research\", \"what's stale\", \"any drift\", \"generate a report\", \"研究库体检\", \"过期档案\", \"论据偏离\", \"孤立主题\"."
}

AlphaGBM Knowledge Base Health Check

Periodic audit of a user's research workspace — stale profiles, drifted theses, orphan pages — with an overall 0-100 score and concrete recommendations.

When to use

User asks for an overview of what's broken / out of date in their KB
User wants to manually trigger a fresh health report
User asks about stale profiles, thesis drift, or orphan items
User mentions "健康体检" / "体检" / "audit" / "health check" / "什么需要更新"

Prerequisites

API Key: env ALPHAGBM_API_KEY (format agbm_xxxx…).
Base URL: default https://alphagbm.zeabur.app. Override via ALPHAGBM_BASE_URL.
Tier requirement for generate: POST /health/generate is Pro-only. Free/Plus users get 403 with upgrade_required: true. Free/Plus users can still read the latest auto-generated weekly report via GET /health.

API Endpoints

All endpoints require Authorization: Bearer $ALPHAGBM_API_KEY.

1. Get latest health report

GET /api/research/health

Response when a report exists:

{
  "success": true,
  "has_report": true,
  "report_date": "2026-04-13",
  "overall_score": 78,
  "stale_profiles": [
    {"ticker": "AAPL", "days_since_update": 21}
  ],
  "thesis_drift": [
    {"ticker": "NVDA", "drift_reason": "Revenue growth now 12% vs 25% when thesis was written"}
  ],
  "orphan_pages": [
    {"ticker": "XYZ", "issue": "Profile has no thesis after 30 days"}
  ],
  "recommendations": [
    {"action": "refresh", "ticker": "AAPL", "reason": "21 days stale"},
    {"action": "review_thesis", "ticker": "NVDA", "reason": "growth decelerated"},
    {"action": "archive", "ticker": "XYZ", "reason": "orphan for 30+ days"}
  ],
  "created_at": "2026-04-13T02:00:00Z"
}

Response when no report yet:

{
  "success": true,
  "has_report": false
}

2. Generate fresh report (Pro-only)

POST /api/research/health/generate

Kicks off immediate audit across the user's profiles + theses + themes. Returns the new report in the same shape as GET.

Tier-blocked response (403):

{
  "success": false,
  "error": "Health check generation is a Pro feature.",
  "upgrade_required": true
}

Free/Plus users still get an auto-generated report weekly (served by GET), just can't trigger on-demand.

Response schema — report

{
  id, report_date,
  stale_profiles,          // [{ticker, days_since_update}]
  thesis_drift,            // [{ticker, drift_reason}]
  orphan_pages,            // [{ticker, issue}]
  overall_score,           // 0-100
  recommendations,         // [{action, ticker, reason}]
  created_at
}

Score band interpretation

Score	Band	Meaning
90-100	Excellent	Nothing urgent
75-89	Good	Minor staleness
60-74	Fair	Several profiles need refresh, some drift
40-59	Poor	Significant drift / orphans
0-39	Critical	KB is mostly out of date

Recommendation action types

refresh — call POST /api/research/profiles/<ticker>/refresh
review_thesis — surface the thesis + updated fundamentals; user decides to edit/close
archive — DELETE /api/research/profiles/<ticker> or delete orphan
create_thesis — profile exists but has no thesis; prompt user to write one

Skills that execute these actions: alphagbm-company-profile, alphagbm-investment-thesis.

Typical Workflow

1. User: "How's my research brain looking?"
   → GET /api/research/health
   → If has_report=false: "No report yet — your first auto-audit runs on <day>.
      Pro users can trigger one now."
   → If has_report=true: present score + top 3 recommendations

2. User (Pro): "Run a fresh health check now"
   → POST /api/research/health/generate
   → Present the new report

3. User (Free/Plus): "Run a health check"
   → POST /api/research/health/generate returns 403
   → Fall back: show the last weekly auto-report via GET, suggest upgrade

4. User: "Fix the stale ones"
   → For each recommendation with action=refresh:
     POST /api/research/profiles/<ticker>/refresh
   → Re-check: GET /api/research/health → show updated score

Output Formatting Tips

When presenting a health report:

Lead with overall score + band (color: green ≥ 75, yellow 60-74, red < 60)
Top 3 recommendations — most actionable first (refresh > review > archive)
Group by category — "3 stale profiles · 1 drifted thesis · 2 orphans" as a chip row
Each ticker actionable — pair the reason with a one-click (or one-prompt) action
Report date prominence — "as of 2026-04-13" so user knows freshness
If no report yet — explain the weekly cadence, offer Pro upgrade for on-demand