repo·evals
· 2026-05-05 ·main@HEAD

follow-builders

zarazhangrui/follow-builders

🛠60 / 100
✅
⚠
🎯
⚠

⚛
→
⚗
→
🧬

🛑
0–29
⚠️
30–49
🛠
50–79
🏭
80–100
▼
60
🛠· 60 / 100
  • ✓5 claims passed, no critical failures
  • ⚠README may claim a license but no LICENSE file exists
  • ◐release_pipeline=1, recently_active=True
  • ✓multilingual_readme=true
  • ⚪static-only eval; live e2e pending

#1👤
#2🎯
#3🧭
#4⇄

25 X builders25 位 X builder+ 6 podcasts+ 6 播客+ 2 blogs+ 2 博客Central pipeline中央 pipeline(scripts/generate-feed)(scripts/generate-feed)feed-{x,podcasts,feed-{x,podcasts,blogs}.jsonblogs}.jsonLocal skill本地 skill+ user prefs+ 用户偏好→ render + translate→ 渲染 + 翻译Telegram / emailTelegram / 邮箱/ in-chat/ chat

Install via OpenClaw or Claude Code skill loaderanyeasy
git clone + agent setup conversationanyeasy
  • 🌐
Zara's central feed pipeline
Pre-fetched + summarized content (the core value)
No API key needed — but it's a single point of failure with no SLA. If her server stops, your digest stops.
Anthropic Claude / OpenClaw
Agent host
Standard agent-side cost
Telegram / Discord / WhatsApp / email
Optional digest delivery channel
Pick one delivery method; in-chat is the no-setup default
· 7
5 1 1
+40
+13
+7
+3
0
-3

6 / 7
passed claim-001

passed claim-002

passed claim-003

passed claim-004

passed claim-005

failed claim-006

untested claim-007

input_contract
output_contract
determinism
idempotence
no_skill_callouts
failure_mode_clarity

workflow_correctness
declared_call_graph
stop_conditions
handoff_points
atom_evidence
error_propagation
partial_failure_handling

  • core user-facing layer untested → capped at 'usable'
  • hybrid-repo rule: archetype 'hybrid-skill' requires end-to-end evaluation of the user-facing layer
  • evidence_completeness='partial' (not portable) → capped at 'usable'

  • only 3/4 critical claims covered

archetype: hybrid-skill→core_layer_tested? False→evidence: partial→recommended: usable→final: usable
ceiling 1 · core user-facing layer untested → capped at 'usable'
ceiling 2 · hybrid-repo rule: archetype 'hybrid-skill' requires end-to-end evaluation of the user-facing layer
ceiling 3 · evidence_completeness='partial' (not portable) → capped at 'usable'
claim-001SKILL.md 完整且 ≥ 200 行（不只是占位）criticalskill-shape● passed
claim-0025 个 prompt 文件真实存在且与 README 一致criticalprompt-coverage● passed
claim-003三种 feed JSON 文件真实存在且对应 README 数字criticalfeed-coverage● passed
claim-004scripts/ 真有 generate-feed + prepare-digest + deliver 三个 Node 脚本highpipeline● passed
claim-005双语支持真实（README + 翻译 prompt）highi18n● passed
claim-006仓库有 LICENSEhighlicensing✕ failed
claim-007端到端 happy path：用户能在 1 次对话内完成订阅criticalend-to-end○ untested

0%
0.00s
0

run-static-checks

2026-05-05
0% — tokens in ? / out ?

run-static-checks

2026-05-05
0% — tokens in ? / out ? 
# follow-builders — final verdict (2026-05-05)

## Repo

- **Name:** zarazhangrui/follow-builders
- **Branch:** main@HEAD · **Stars:** 3,689
- **Archetype:** hybrid-skill · **Layer:** molecule

## What was evaluated

| Claim | Status | Notes |
|---|---|---|
| 001 SKILL.md ≥ 200 lines | passed | 466-line SKILL.md with frontmatter |
| 002 5 prompt files | passed | All present, 1.0–2.6 KB each |
| 003 source list = README | passed | config/default-sources.json: 25 X + 6 podcasts + 2 blogs |
| 004 Node.js pipeline | passed | generate-feed (38 KB) + prepare-digest (5 KB) + deliver (7 KB) |
| 005 bilingual | passed | README.zh-CN.md + prompts/translate.md |
| 006 LICENSE | **failed** | HTTP 404 |
| 007 live e2e | untested | needs Claude Code session + delivery channel |

## Real findings

1. **No LICENSE** — same gap as `codebase-to-course`. Author-level
   pattern. Easy upstream fix.

2. **The "no API keys needed" architecture is clever.** The user-facing
   skill is light because Zara runs a central Node.js pipeline that
   pre-fetches and summarizes everything. Trade-off: every user is
   tightly coupled to her server uptime — single point of failure
   with no documented SLA or fallback.

3. **466-line SKILL.md is unusually thorough.** Most skills are
   100-200 lines. The depth here suggests serious investment in
   making the conversational setup work without config files.

4. **38 KB `generate-feed.js`** — this is real engineering, not a
   demo. The maintainer has invested in the central pipeline.

## Why not higher

- LICENSE missing (-5 score, claim-006 failed).
- Live e2e not logged (molecule layer cap; central pipeline single
  point of failure means a live test is more important than usual).

## Path forward

1. Add LICENSE to repo root.
2. Run a real setup + delivery scenario, ideally with Telegram bot
   token, log first-digest delivery.
3. Test "central server unreachable" scenario; verify error surfaces
   to the user (not silent skip).
4. Update claim-006 + claim-007.