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 agentmods add commands/nirelbaz/promptpit/refresh-knowledgegit clone --depth 1 https://github.com/nirelbaz/promptpitWhat 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 | $0.00000 | $0.00501 |
| Opus 5 | $0.00000 | $0.00251 |
| Sonnet 5 | $0.00000 | $0.00100 |
| Haiku 4.5 | $0.00000 | $0.00050 |
Grade A, and why
refresh-knowledge 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 2d 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 — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Read the AI Stack Expert agent definition at .claude/agents/ai-stack-expert.md to understand your role, expertise, and methodology. Then execute the following refresh workflow.
Scope
$ARGUMENTS
If a specific tool name was provided above (e.g., "cursor"), refresh only that tool's knowledge file. Otherwise, refresh all tools in docs/knowledge/.
Workflow
Step 0: Inventory
Read all files in docs/knowledge/ and sort by last-verified date (oldest first). This is your refresh priority order.
Step 1: Per-Tool Refresh
For each tool (or the specified tool):
- Read current knowledge file from
docs/knowledge/<tool>.md - Check freshness — if
last-verifiedis less than 7 days old and no specific tool was requested, skip it - Research current state using WebSearch and WebFetch:
- Search the tool's official documentation (URLs in the
doc-urlsfrontmatter) - Check for recent changelog entries, blog posts, or release notes
- Search for configuration documentation: file paths, formats, supported features
- Search for MCP server support, agent support, rules/instructions support
- Search for any cross-tool reading behavior (does it read AGENTS.md, .mcp.json, etc.?)
- Search the tool's official documentation (URLs in the
- Update the knowledge file with verified findings:
- Fill in or update all sections: Configuration, Cross-Tool Reading, Behavior, Ecosystem, Edge Cases
- For tools with
status: adapter-exists, update the Promptpit Gaps section by comparing what the tool supports vs what the adapter implements - Update
last-verifiedto today's date - Add or update
doc-urlswith any new documentation sources discovered
- Do NOT modify any source code — this command only updates knowledge files
Step 2: Summary
After refreshing, print a summary:
## Knowledge Refresh Summary — YYYY-MM-DD
### Refreshed
- <tool>: <what changed or was verified>
### Skipped (recently verified)
- <tool>: last verified YYYY-MM-DD
### Breaking Changes Detected
- <any changes that affect promptpit adapters>
### New Tools Discovered
- <any new AI coding tools worth tracking>
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.
- 2d ago First seen · 52 lines · 0 tokens per session scan A 2b4cdaf49178
refresh-knowledge is a command published in the GitHub repository nirelbaz/promptpit (6 stars, last pushed 1mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 501 tokens. 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 2026-08-31.
Other commands, from other repositories
fix-layout
Command "fix-layout" from OpenRaiser/PaperFit, covering /fix-layout — 启动完整 vto 排版优化, 用户入口说明, 工具调用约定, 用法 and 直接执行入口.
paperfit
作用: /paperfit 是 Claude Code 中的 PaperFit 主入口。用户可以直接用自然语言描述任务，例如排版分析、完整修复、模板迁移、长度调整、局部表格修复、视觉检查或状态查询。Agent 负责解析意图、识别主 .tex、判断是否需要画像、并自动进入对应的 PaperFit 子流程。.
paperfit-undo
作用: 恢复最近一次 PaperFit 自动写回前的备份版本，优先回滚主 .tex，必要时一并恢复 data/state.json。.
show-status
作用: 显示当前 PaperFit 任务的运行状态、缺陷消除进度、视觉优先级、修复计划摘要和下一步行动；这是当前默认的“摘要 / 解释”入口。.
adjust-length
作用: 尝试通过排版微调或受控语义改写逼近目标页数。它是专家快捷入口；普通自然语言如“把正文压到 8 页，尽量不要改语义”也应能触发同类任务。.
repair-table
作用: 针对指定表格或当前论文中的表格问题执行修复闭环。它是专家快捷入口；普通自然语言如“修 Table 2 太挤的问题”也应能触发同类任务。.