# Oracle

> Use when bundled prompts and files need a second model's review.

Oracle second-model review: bundle prompts/files, debug, refactor, design.

## Facts

- Type: Agent skill
- Discipline: Engineering
- Author: Peter Steinberger (https://github.com/steipete)
- Upstream: https://github.com/steipete/agent-scripts/tree/main/skills/oracle
- Source file: https://raw.githubusercontent.com/steipete/agent-scripts/main/skills/oracle/SKILL.md (8400 bytes, sha256:e08d6cc6a499db075e710cc9121efa2ad9f45cd1cbfffc1bc346f83c5fc6832f)
- Licence: MIT
- Upstream copy last changed: 2026-09-17
- Status: listed, not yet ruled on by the curator

## Use it

I want to use the Oracle skill by Peter Steinberger.
This is what it does, in the author's own words: Oracle second-model review: bundle prompts/files, debug, refactor, design.
Download the whole folder at https://github.com/steipete/agent-scripts/tree/main/skills/oracle, SKILL.md and every file beside it, into your skills directory: ~/.claude/skills/oracle/ for Claude Code, ~/.agents/skills/oracle/ for Codex, ~/.cursor/skills/oracle/ for Cursor. Get it from the author's repo, not from startupskills.dev.
Then tell me the path you installed it to and how to invoke it.

Nothing here is vendored: install from the author's repo, not from this site. A verbatim copy of the author's file follows for reading, under the licence reproduced after it; the current file is in the author's repo.

## Links

- Page: https://startupskills.dev/steipete/oracle
- Author: https://startupskills.dev/steipete
- Index: https://startupskills.dev/llms.txt

## The file itself

---
name: oracle
description: "Oracle second-model review: bundle prompts/files, debug, refactor, design."
---

# Oracle (CLI) — 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 outputs as advisory: verify against the codebase + tests.

## Main use case (browser, GPT‑5.5 Pro)

Default workflow here: `--engine browser` with GPT‑5.5 Pro in ChatGPT. This is the “human in the loop” path: it can take ~10 minutes to ~1 hour; expect a stored session you can reattach to.

Recommended defaults:
- Engine: browser (`--engine browser`)
- Model: GPT‑5.5 Pro (either `--model gpt-5.5-pro` or a ChatGPT picker label like `--model "5.5 Pro"`)
- Attachments: directories/globs + excludes; avoid secrets.

## Golden path (fast + reliable)

1. Pick a tight file set (fewest files that still contain the truth).
2. Preview what you’re about to send (`--dry-run` + `--files-report` when needed).
3. Run in browser mode for the usual GPT‑5.5 Pro ChatGPT 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)

- Show help (once/session):
  - `npx -y @steipete/oracle --help`

- Preview (no tokens):
  - `npx -y @steipete/oracle --dry-run summary -p "<task>" --file "src/**" --file "!**/*.test.*"`
  - `npx -y @steipete/oracle --dry-run full -p "<task>" --file "src/**"`

- Token/cost sanity:
  - `npx -y @steipete/oracle --dry-run summary --files-report -p "<task>" --file "src/**"`

- Startup/perf trace:
  - `npx -y @steipete/oracle --perf-trace --perf-trace-path /tmp/oracle-perf.json --dry-run summary -p "<task>" --file "src/**"`
  - Use when CLI startup or time-to-first-output feels slow; inspect `first-output` and `exit`.

- Browser run (main path; long-running is normal):
  - `npx -y @steipete/oracle --engine browser --model gpt-5.5-pro -p "<task>" --file "src/**"`

- Manual paste fallback (assemble bundle, copy to clipboard):
  - `npx -y @steipete/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/**"` (directory glob)
  - `--file src/index.ts` (literal file)
  - `--file docs --file README.md` (literal directory + file)

- Exclude (prefix with `!`):
  - `--file "src/**" --file "!src/**/*.test.ts" --file "!**/*.snap"`

- Defaults (important behavior from the implementation):
  - Default-ignored dirs: `node_modules`, `dist`, `coverage`, `.git`, `.turbo`, `.next`, `build`, `tmp` (skipped unless you explicitly pass them as literal dirs/files).
  - Honors `.gitignore` when expanding globs.
  - Does not follow symlinks (glob expansion uses `followSymbolicLinks: false`).
  - Dotfiles are filtered unless you explicitly opt in with a pattern that includes a dot-segment (e.g. `--file ".github/**"`).
  - Default cap: files > 1 MB are rejected unless you raise `ORACLE_MAX_FILE_SIZE_BYTES` or `maxFileSizeBytes` in `~/.oracle/config.json`.

## Budget + observability

- Target: keep total input under ~196k tokens.
- Use `--files-report` (and/or `--dry-run json`) to spot the token hogs before spending.
- Use `--perf-trace` / `ORACLE_PERF_TRACE=1` for startup and first-output timing. Traces redact prompts, tokens, keys, cookies, and inline cookie payloads; detached API children write a session-suffixed sidecar trace.
- If you need hidden/advanced knobs: `npx -y @steipete/oracle --help --verbose`.

## Engines (API vs browser)

- Auto-pick: uses `api` when `OPENAI_API_KEY` is set, otherwise `browser`.
- Browser engine supports GPT + Gemini only; use `--engine api` for Claude/Grok/Codex or multi-model runs.
- **API runs require explicit user consent** before starting because they incur usage costs.
- Browser attachments:
  - `--browser-attachments auto|never|always` (auto pastes inline up to ~60k chars then uploads).
  - Add `--browser-bundle-files --browser-bundle-format zip` to upload many text files as one ZIP while preserving file names.
- Remote browser host (signed-in machine runs automation):
  - 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/**"`

## API preflight

- API runs require explicit user consent and cost money.
- Before API runs, check provider readiness without printing secrets:
  - `oracle doctor --providers --models gpt-5.4,claude-4.6-sonnet,gemini-3-pro`
  - `oracle --preflight --models gpt-5.4,gemini-3-pro`
  - `oracle --route --model gpt-5.4`
- If the user wants first-party OpenAI, pass `--provider openai` or `--no-azure`. This prevents exported Azure env/config from hijacking the route:
  - `oracle --provider openai --engine api --model gpt-5.5-pro ...`
- For advisory multi-model panels where partial success is useful, use `--allow-partial --write-output <path>` so successful model files and the `<stem>.oracle.json` manifest are easy to recover:
  - `oracle --models gpt-5.4,claude-4.6-sonnet,gemini-3-pro --allow-partial --write-output /tmp/panel.md -p "<task>"`
- `--timeout 10m` is the normal user-facing API deadline; Oracle derives the HTTP transport timeout unless `--http-timeout` is explicitly set.
- If the exported `OPENAI_API_KEY` is invalid and the user wants their personal OpenAI key, use `$one-password` in one persistent tmux session. Known item: `API Key - OpenAI - Personal`, field `api_key`. Inject only into the single Oracle command; never print the key:
  - `OPENAI_API_KEY="$(op item get 'API Key - OpenAI - Personal' --account my.1password.com --fields label=api_key --reveal)" oracle --provider openai --engine api --model gpt-5.5-pro ...`
- For debugging Oracle itself, prefer the local checkout after pulling `~/Projects/oracle`:
  - `pnpm -C ~/Projects/oracle run build`
  - `node ~/Projects/oracle/dist/scripts/run-cli.js ...`

## Sessions + slugs (don’t lose work)

- Stored under `~/.oracle/sessions` (override with `ORACLE_HOME_DIR`).
- Browser runs save durable files under `~/.oracle/sessions/<id>/artifacts/`, including `transcript.md`, Deep Research reports, and downloaded ChatGPT-generated images when available.
- Runs may detach or take a long time (browser/API + GPT‑5.5 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.
- CLI guardrails: root runs without a prompt exit nonzero; `--dry-run` conflicts with `--render` / `--render-markdown`; Ctrl-C exits foreground API runs with code 130 while browser cleanup/reattach still runs.

## 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, dependency boundaries).
- Exact question + what you tried + the error text (verbatim).
- Constraints (“don’t change X”, “must keep public API”, “perf budget”, etc).
- Desired output (“return patch plan + tests”, “list risky assumptions”, “give 3 options with tradeoffs”).

### “Exhaustive prompt” pattern (for later restoration)

When you know this will be a long investigation, write a prompt that can stand alone later:
- Top: 6–30 sentence project briefing + current goal.
- Middle: concrete repro steps + exact errors + what you already tried.
- Bottom: attach *all* context files needed so a fresh model can fully understand (entrypoints, configs, key modules, docs).

If you need to reproduce the same context later, re-run with the same prompt + `--file …` set (Oracle runs are one-shot; the model doesn’t remember prior runs).

## Safety

- Don’t attach secrets by default (`.env`, key files, auth tokens). Redact aggressively; share only what’s required.
- Prefer “just enough context”: fewer files + better prompt beats whole-repo dumps.

## Licence of the file above

The file above is https://raw.githubusercontent.com/steipete/agent-scripts/main/skills/oracle/SKILL.md, by Peter Steinberger, reproduced without changes.
- Source: https://github.com/steipete/agent-scripts/tree/main/skills/oracle
- Copyright: Copyright (c) 2026 Peter Steinberger
- Licence: MIT, https://raw.githubusercontent.com/steipete/agent-scripts/main/LICENSE

The repo's licence file, verbatim from https://raw.githubusercontent.com/steipete/agent-scripts/main/LICENSE:

MIT License

Copyright (c) 2026 Peter Steinberger

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.