Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add ChanningLua/prax-agent --skill article-batchgit clone --depth 1 https://github.com/ChanningLua/prax-agentWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/channinglua/prax-agent/article-batch)<a href="https://agentmods.dev/skills/channinglua/prax-agent/article-batch"><img src="https://agentmods.dev/badge/skills/channinglua/prax-agent/article-batch/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/channinglua/prax-agent/article-batch"><img src="https://agentmods.dev/badge/skills/channinglua/prax-agent/article-batch.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00030 | $0.01531 |
| Opus 5.5 | $0.00012 | $0.00612 |
| Sonnet 5.5 | $0.00006 | $0.00306 |
| Haiku 4.5 | $0.00003 | $0.00153 |
Grade A, and why
article-batch scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 142 lines — stays where its author put it; the contents beside it link to each section on GitHub.
十篇文章批处理
这个技能负责吞吐,不降低 hotspot-article 的证据和质量门。十篇文章使用十个独立工作区,任何一篇失败都不能污染其他文章。
目标与边界
- 默认
count=10,允许 2–10; - 默认组合:
event 30% / demand 50% / evergreen 20%; - 每篇都要有完整业务场景和七段技术到业务桥;
- 每篇先过题目重量门;小版本、小功能和局部技术点直接降级,不占十篇精选名额;
- 每篇独立维护来源、实测、事实核验和质量报告;
- 不为了凑十篇复制角度、来源或段落;
- 不自动发布。当天未通过的文章保留草稿和失败原因。
Step 1:检查开源适配器
python3 -m prax.article_integrations probe --format json
按可用性使用:
- TrendRadar:聚合公开热榜与 RSS,读取
TRENDRADAR_OUTPUT; - AutoCLI:抓取需要 Chrome 登录态的平台;
- Crawl4AI:把公开文章转成研究用 Markdown;缺失时使用遵守
robots.txt的 HTTP 回退; - GPT Researcher:并行生成研究线索;需要本地包、模型和检索凭据;
- STORM:读取
STORM_OUTPUT中已有报告,补充视角和追问。
外部研究报告不是事实来源。必须回到其中列出的原始 URL。
Step 2:建立至少 12 个候选
十篇批次至少准备 12 个候选,避免一个题失败后整批停住。候选写入:
.prax/vault/article-batches/<DATE>/topic-candidates.json
候选必须符合 hotspot-article 的 business_case 和 business_bridge schema。只复述发布会或产品功能的题目不得进入批次。
Step 3:创建十个隔离工作区
python3 -m prax.article_batch \
".prax/vault/article-batches/<DATE>/topic-candidates.json" \
--count 10 \
--date "<DATE>" \
--out-root ".prax/vault/article-batches/<DATE>/run"
生成:
run/
├── batch-manifest.json
├── calibration-gate.json
├── 01-<topic>/
│ ├── brief.md
│ ├── business-bridge.json
│ ├── research-plan.json
│ └── status.json
└── 10-<topic>/...
batch-manifest.json 没有 complete: true 时,不开始写作;先补候选。重复运行必须显式使用 --resume,避免误覆盖批次状态。
Step 4:先用一篇做校准门
批次创建后只推进 calibration-pilot,其他目录保持 waiting-for-calibration。首篇必须完整走到来源、实测、事实审校、去 AI 腔编辑和质量门,再把实际暴露的问题写入 calibration-gate.json:
- 机器分在哪些地方虚高或误判;
- 哪些来源只进研究池、没有真正进入正文;
- 计划验证与实际运行之间还差什么;
- 哪些段落虽然过门,读起来仍像模板;
- 题目是否足以改变一个真实业务决策,还是被篇幅硬撑成“大文章”;
- 已经修改了哪些 prompt、规则或 hard gate。
只有首篇 hard gates 通过、编辑七项均至少 4/5、证据明确写出局限,并且上述改动已经落地,才把 status 改为 released、remaining_slots_released 改为 true。不能先批量生成十篇再统一返工。
Step 5:校准后按两条并发泳道推进
不要十篇同时抢浏览器和检索限额。最多同时研究两篇:
- 泳道 A 研究第 1 篇时,泳道 B 研究第 2 篇;
- A 进入写作后,再把第 3 篇放入研究;
- 每完成一个阶段就更新该目录的
status.json; - 单篇失败只标记
blocked,继续下一篇。
建议状态:
planned → researching → outlined → drafting → fact-checking → quality-gate → ready
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 142 lines · 30 tokens per session scan A 3203ee09a90a
article-batch is a skill published in the GitHub repository ChanningLua/prax-agent (272 stars, last pushed 1mo ago), licensed MIT. It adds 30 tokens to every session and 1,531 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-12.
Other skills, from other repositories
together-ai-inference
Serverless inference, fine-tuning, embeddings, image generation, and batch processing on 200+ open-source models via an OpenAI-compatible API. Use when you need fast, cost-effective access to open-source LLMs without managing infrastructure.
ecommerce-full-pipeline
An e-commerce workflow assistant covering product discovery, 1688 product collection, listing on multiple platforms, promotional copy, and short-video creation. 1688 is a Chinese online wholesale marketplace.
cis-aws-compute-8.2
Ensure Batch roles are configured for cross-service confused deputy prevention.
cis-aws-compute-8.1
Ensure AWS Batch is configured with AWS CloudWatch Logs.
data-batch-processing
Use this skill when designing batch processing with Hive, Spark SQL, Pig, or HQL. This skill enforces: Hive metastore management, Spark SQL Catalyst optimization, file format selection (Parquet, ORC, Avro), partitioning and bucketing strategies, query tuning with statistics and dynamic partition pruning. Do NOT use…
servicenow-import-attachment-batch
ServiceNow data-ingest, file, and batched-REST surface — push rows through Import Sets (single and bulk), upload/get/delete record attachments, and bundle many REST calls into one batch request via the servicenow-api MCP server. Use when the agent must load external data through a transform-mapped import set, attach…