oracle

作者: firecrawl

使用oracle CLI的最佳实践(提示词与文件打包、引擎、会话及文件附件模式)。

npx skills add https://github.com/firecrawl/openclaw --skill oracle

oracle — best use

Oracle bundles your prompt + selected files into one “one-shot” request so another model can answer with real repo context (API or browser automation). Treat output as advisory: verify against code + tests.

Main use case (browser, GPT‑5.2 Pro)

Default workflow here: --engine browser with GPT‑5.2 Pro in ChatGPT. This is the common “long think” path: ~10 minutes to ~1 hour is normal; expect a stored session you can reattach to.

Recommended defaults:

  • Engine: browser (--engine browser)
  • Model: GPT‑5.2 Pro (--model gpt-5.2-pro or --model "5.2 Pro")

Golden path

  1. Pick a tight file set (fewest files that still contain the truth).
  2. Preview payload + token spend (--dry-run + --files-report).
  3. Use browser mode for the usual GPT‑5.2 Pro workflow; use API only when you explicitly want it.
  4. If the run detaches/timeouts: reattach to the stored session (don’t re-run).

Commands (preferred)

  • Help:

    • oracle --help
    • If the binary isn’t installed: npx -y @steipete/oracle --help (avoid pnpx here; sqlite bindings).
  • Preview (no tokens):

    • oracle --dry-run summary -p "<task>" --file "src/**" --file "!**/*.test.*"
    • oracle --dry-run full -p "<task>" --file "src/**"
  • Token sanity:

    • oracle --dry-run summary --files-report -p "<task>" --file "src/**"
  • Browser run (main path; long-running is normal):

    • oracle --engine browser --model gpt-5.2-pro -p "<task>" --file "src/**"
  • Manual paste fallback:

    • oracle --render --copy -p "<task>" --file "src/**"
    • Note: --copy is a hidden alias for --copy-markdown.

Attaching files (--file)

--file accepts files, directories, and globs. You can pass it multiple times; entries can be comma-separated.

  • Include:

    • --file "src/**"
    • --file src/index.ts
    • --file docs --file README.md
  • Exclude:

    • --file "src/**" --file "!src/**/*.test.ts" --file "!**/*.snap"
  • Defaults (implementation behavior):

    • Default-ignored dirs: node_modules, dist, coverage, .git, .turbo, .next, build, tmp (skipped unless explicitly passed as literal dirs/files).
    • Honors .gitignore when expanding globs.
    • Does not follow symlinks.
    • Dotfiles filtered unless opted in via pattern (e.g. --file ".github/**").
    • Files > 1 MB rejected.

Engines (API vs browser)

  • Auto-pick: api when OPENAI_API_KEY is set; otherwise browser.
  • Browser supports GPT + Gemini only; use --engine api for Claude/Grok/Codex or multi-model runs.
  • Browser attachments:
    • --browser-attachments auto|never|always (auto pastes inline up to ~60k chars then uploads).
  • Remote browser host:
    • Host: oracle serve --host 0.0.0.0 --port 9473 --token <secret>
    • Client: oracle --engine browser --remote-host <host:port> --remote-token <secret> -p "<task>" --file "src/**"

Sessions + slugs

  • Stored under ~/.oracle/sessions (override with ORACLE_HOME_DIR).
  • Runs may detach or take a long time (browser + GPT‑5.2 Pro often does). If the CLI times out: don’t re-run; reattach.
    • List: oracle status --hours 72
    • Attach: oracle session <id> --render
  • Use --slug "<3-5 words>" to keep session IDs readable.
  • Duplicate prompt guard exists; use --force only when you truly want a fresh run.

Prompt template (high signal)

Oracle starts with zero project knowledge. Assume the model cannot infer your stack, build tooling, conventions, or “obvious” paths. Include:

  • Project briefing (stack + build/test commands + platform constraints).
  • “Where things live” (key directories, entrypoints, config files, boundaries).
  • Exact question + what you tried + the error text (verbatim).
  • Constraints (“don’t change X”, “must keep public API”, etc).
  • Desired output (“return patch plan + tests”, “give 3 options with tradeoffs”).

Safety

  • Don’t attach secrets by default (.env, key files, auth tokens). Redact aggressively; share only what’s required.

“Exhaustive prompt” restoration pattern

For long investigations, write a standalone prompt + file set so you can rerun days later:

  • 6–30 sentence project briefing + the goal.
  • Repro steps + exact errors + what you tried.
  • Attach all context files needed (entrypoints, configs, key modules, docs).

Oracle runs are one-shot; the model doesn’t remember prior runs. “Restoring context” means re-running with the same prompt + --file … set (or reattaching a still-running stored session).

来自 firecrawl 的更多技能

firecrawl-research-index
firecrawl
使用 Firecrawl Research 查找回答研究查询的论文,采用语义搜索、语义与结构扩展以及正文内验证。对于任何文献查找/论文检索任务——无论是单篇论文查找还是完整的多篇论文集合——始终使用此技能。
data-analysisresearchweb-scraping
pinecone
firecrawl
面向生产级AI应用的托管向量数据库。全托管、自动扩缩容,支持混合搜索(稠密+稀疏)、元数据过滤和命名空间。
wpds
firecrawl
在构建利用WordPress设计系统(WPDS)及其组件、令牌、模式等的用户界面时使用。
audiocraft-audio-generation
firecrawl
用于音频生成的PyTorch库,包括文本生成音乐(MusicGen)和文本生成声音(AudioGen)。当需要从文本生成音乐时使用…
skypilot-multi-cloud-orchestration
firecrawl
跨多云编排机器学习工作负载,自动优化成本。当您需要在多个云上运行训练或批处理作业时使用,利用…
firecrawl-seo-audit
firecrawl
使用 Firecrawl 对网站进行 SEO 审计。适用于用户要求进行 SEO 审计、元数据和标题审查、站点地图/网站结构分析、关键词机会、竞争对手 SERP 对比,或优先搜索优化建议的场景。
data-analysisresearchweb-scraping
gh-issues
firecrawl
获取GitHub议题,生成子代理来实施修复并开启拉取请求,然后监控并处理PR审查评论。用法:/gh-issues [owner/repo] [--label…]
tmux
firecrawl
通过发送按键和抓取窗格输出来远程控制tmux会话,以操作交互式命令行界面。