oracle-chatgpt
Run exactly one arbitrary prompt through ChatGPT via Oracle browser mode and return the captured response without reinterpretation.
Skill by dceoy
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/dceoy-oracle-chatgpt
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: productivity, agent-building
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/oracle-chatgpt
Tools it pre-approves 4
From allowed-tools in SKILL.md: the agent may use these without asking you each time.
Bash(oracle:*),Bash(which:*),Bash(mktemp:*),Write
What's inside
- Skill name
oracle-chatgpt- Description
- Run exactly one arbitrary prompt through ChatGPT via Oracle browser mode and return the captured response without reinterpretation. Use when the user explicitly wants a generic ChatGPT-via-Oracle consult without task-specific orchestration.
- License
- AGPL-3.0
- Allowed tools
- Bash(oracle:*),, Bash(which:*),, Bash(mktemp:*),, Write
- Instructions
- 142 lines in SKILL.md · view on GitHub
Before installing
- Pre-approves some shell commands
Files (1)
- SKILL.md
SKILL.md (first lines)
# Oracle ChatGPT Run exactly one arbitrary prompt through ChatGPT using Oracle browser mode. Treat the prompt as opaque data: preserve its content, let Oracle own browser/session and remote routing, wait for the run to finish, and return the captured ChatGPT response without semantic rewriting. ## When to Use Use this skill when the user explicitly wants to: - send an arbitrary prompt to ChatGPT through Oracle; - reuse Oracle's browser/session or remote-browser routing without adding task-specific workflow logic; or - obtain ChatGPT's response verbatim for another agent workflow. Do not use this skill to resolve GitHub targets, gather repository context, construct PR/issue-specific prompts, or post results to external services. ## Prerequisites Require: - `oracle` in `PATH`; - an authenticated ChatGPT browser session usable by Oracle; and - `GPT-5.6 Sol` available to Oracle browser mode. Check Oracle with: ```bash which oracle ``` Stop if a required prerequisite is unavailable. ## Input Contract Accept exactly one prompt from the caller. Do not semantically modify the prompt: do not prepend, append, summarize, reinterpret, escape into executable shell text, or otherwise augment its content. Do not attach repository files or other context by default. Treat quotes, newlines, backticks, command substitutions such as `$()`, variable syntax such as `$HOME`, and other shell-significant characters as prompt data only. Oracle's `-p -` stdin transport (see Prompt Transport) trims leading/trailing whitespace from the prompt and rejects a prompt that is empty after trimming; that trimming is Oracle's behavior, not a semantic modification this skill introduces. Do not compensate for it by rewriting the prompt or switching to another transport — a whitespace-only prompt fails under Failure and Output like any other Oracle rejection. ## Prompt Transport Pass the prompt to Oracle through stdin with `-p -`. Do not place arbitrary prompt text directly in a shell command, use `eval`, or construct an `echo`, `printf`, heredoc, or similar shell expression from the prompt. `-p -` is a shell-safe transport for arbitrary multiline, quote-, and shell-metacharacter-containing prompt data, but it is not byte-exact: Oracle trims outer (leading/trailing) whitespace from stdin before use and rejects the prompt if nothing remains after trimming. Internal newlines and all other prompt content pass through unchanged. Do not add a wrapper, alternate CLI flag, or other transport to preserve outer whitespace; accept Oracle's trimming behavior as part of this transport. Prefer the runtime's direct stdin/data channel when it can supply the prompt bytes to the Oracle process without shell interpolation. If the runtime cannot bind stdin directly: 1. Run `mktemp` and capture the absolute path it prints. 2. Use the runtime's file-writing capability to write exactly the caller's prompt to that exact path as data. 3. Redirect that exact absolute path to Oracle stdin. Substitute the literal path returned by `mktemp`; do not rely on a shell variable or other shell state surviving between tool calls. 4. After Oracle exits, use the file-writing capability to overwrite the same temporary file with empty content. Do this on both success and failure paths when possible; the operating system may later remove the empty temporary entry. Only the generated temporary-file path may appear in the shell command. The prompt itself must not. Treat the `mktemp` output as an opaque path and shell-quote that literal path when using redirection. ## Oracle Routing Keep Oracle's native browser/session and remote-host routing intact. Do not add `--remote-host` or `--remote-token`, print remote credentials, or reproduce Oracle's configuration-precedence logic in this skill. Oracle may resolve supported local or remote browser settings from its own configuration and environment, including `ORACLE_REMOTE_HOST` and `ORACLE_REMOTE_TOKEN`. Treat Oracle routing, browser, authentication, or model-selection failures as failures rather than switching execution paths. ## Run Invoke Oracle in the foreground and wait for it to complete. When using the temporary-file transport, replace `/absolute/path/from-mktemp` below with the exact absolute path returned by the preceding `mktemp` tool call: ```bash oracle \ --engine browser \ --model gpt-5.6-sol \ --browser-thinking-time high \ -p - < '/absolute/path/from-mktemp' ``` Do not use `$prompt_file` or another shell variable to carry the path across separate tool calls. When the runtime supplies stdin directly, invoke the same argv without shell redirection and provide the prompt through that stdin channel. Do not add `--file`, switch engines or models, or retry with a modified prompt. ## Failure and Output Fail closed. If Oracle exits non-zero or reports a browser, authentication, routing, or model-selection failure, return that failure and stop. Do not fall back to API mode, another model, another browser path, or the calling agent's own answer. On success, return the captured ChatGPT response without summarizing, rewriting, correcting, or replacing it. Keep calling-agent commentary separate only when required to report an execution failure. ## Examples ### Simple prompt Prompt data:
From dceoy/ai-coding-agent-skills, AGPL-3.0.
More from dceoy/ai-coding-agent-skills
- 0.611score
gas-drive-html-server
dceoy
Create or update an independent Google Apps Script Web App that indexes and serves HTML files from a specified Google Drive folder, including its subfolders.
SkillAFreshness: A★ 4Claude Code +9 - 0.611score
gas-html-artifact
dceoy
Deploy an existing HTML artifact to Google Apps Script HTML Service with Google's clasp CLI, preserving compatible bytes or applying reviewed compatibility…
SkillAFreshness: A★ 4Claude Code +9 - 0.608score
x-timeline
dceoy
Read an authenticated X Following or For You timeline through agent-browser without APIs or engagement actions.
SkillAFreshness: A★ 4Claude Code +9 - 0.603score
local-qa
dceoy
Run local QA including formatting, linting, and testing for the repository.
SkillAFreshness: A★ 4Claude Code +9 - 0.571score
github-productivity
dceoy
Collect, retain, normalize, and analyze GitHub organization repository/PR/review/timeline/commit data through `gh api` into an organization-week delivery…
SkillBFreshness: B★ 4Claude Code +9 - 0.571score
parameterized-tests
dceoy
Enforce native parameterized or table-driven tests when adding or modifying unit tests that exercise the same behavior with multiple data cases.
SkillBFreshness: B★ 4Claude Code +9 - 0.569score
simplify-codebase
dceoy
Reduce a repository's maintenance surface across the whole codebase with KISS, DRY, and YAGNI while preserving required behavior.
SkillBFreshness: B★ 4Claude Code +9 - 0.569score
claude-agent-converter
dceoy
Convert Claude Code subagents (.claude/agents/*.md) to standard Agent Skills (skills/*/SKILL.md).
SkillBFreshness: B★ 4Claude Code +9 - 0.564score
commit
dceoy
Create a single git commit from current staged and unstaged changes with an appropriate message.
SkillBFreshness: B★ 4Claude Code +9 - 0.564score
commit-push-pr
dceoy
Commit staged and unstaged changes, push the branch to origin, and open a pull request using the GitHub CLI.
SkillBFreshness: B★ 4Claude Code +9 - 0.563score
clean-gone-branches
dceoy
Remove local git branches whose configured upstream is marked [gone], including associated worktrees.
SkillBFreshness: B★ 4Claude Code +9 - 0.519score
claude-command-converter
dceoy
Convert Claude Code commands (.claude/commands/*.md) to standard Agent Skills (skills/*/SKILL.md).
SkillCFreshness: C★ 4Claude Code +9
Signals
- Stars
- 4
- Forks
- 1
- Open issues
- 2
- Last commit (this folder)
- Aug 12, 2026 (2 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- AGPL-3.0
- Repository
- github.com/dceoy/ai-coding-agent-skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from dceoy
- 0.611score
gas-drive-html-server
dceoy
Create or update an independent Google Apps Script Web App that indexes and serves HTML files from a specified Google Drive folder, including its subfolders.
SkillAFreshness: A★ 4Claude Code +9 - 0.611score
gas-html-artifact
dceoy
Deploy an existing HTML artifact to Google Apps Script HTML Service with Google's clasp CLI, preserving compatible bytes or applying reviewed compatibility…
SkillAFreshness: A★ 4Claude Code +9 - 0.608score
x-timeline
dceoy
Read an authenticated X Following or For You timeline through agent-browser without APIs or engagement actions.
SkillAFreshness: A★ 4Claude Code +9 - 0.603score
local-qa
dceoy
Run local QA including formatting, linting, and testing for the repository.
SkillAFreshness: A★ 4Claude Code +9 - 0.571score
github-productivity
dceoy
Collect, retain, normalize, and analyze GitHub organization repository/PR/review/timeline/commit data through `gh api` into an organization-week delivery…
SkillBFreshness: B★ 4Claude Code +9 - 0.571score
parameterized-tests
dceoy
Enforce native parameterized or table-driven tests when adding or modifying unit tests that exercise the same behavior with multiple data cases.
SkillBFreshness: B★ 4Claude Code +9
Cite this page
oracle-chatgpt skill. Agent Mart. https://agentmart.dev/skills/dceoy-oracle-chatgpt (last updated 2026-10-24)