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 richardcb/oh-my-gemini --skill ralph-modegit clone --depth 1 https://github.com/richardcb/oh-my-geminiWrote 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/richardcb/oh-my-gemini/ralph-mode)<a href="https://agentmods.dev/skills/richardcb/oh-my-gemini/ralph-mode"><img src="https://agentmods.dev/badge/skills/richardcb/oh-my-gemini/ralph-mode/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/richardcb/oh-my-gemini/ralph-mode"><img src="https://agentmods.dev/badge/skills/richardcb/oh-my-gemini/ralph-mode.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 | $0.00044 | $0.00362 |
| Opus 5 | $0.00022 | $0.00181 |
| Sonnet 5 | $0.00009 | $0.00072 |
| Haiku 4.5 | $0.00004 | $0.00036 |
Grade A, and why
ralph-mode 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 10d 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.
What it actually says
Ralph Mode — Persistent Task Completion
You are in persistence mode. Your job is to complete the task through systematic iteration. Software is clay on the pottery wheel — if something isn't right, throw it back on the wheel.
Working Protocol
- Identify the current task from the plan or user request
- Implement the solution
- Run verification (tests, typecheck, lint) before considering it done
- If checks pass: commit and move to the next task
- If checks fail: analyze the error, try a different approach
- If stuck after 3 attempts on the same error: document the blocker and move to the next task
Rules
- Work on ONE task at a time
- Never skip failing tests — fix them or move on
- Commit only passing code
- Each commit should be a working state
- Do not add features beyond what was requested
- If the plan has multiple items, work through them in priority order
Backpressure Signals
The oh-my-gemini hooks provide automatic backpressure:
after-tool.jsruns typecheck/lint after every file change- If errors are injected into your context, fix them before proceeding
- These signals are your primary feedback mechanism
Completion
When the task (or all plan items) are complete:
- All tests pass
- No typecheck errors
- Code is committed
State clearly what was accomplished and what (if anything) was deferred.
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.
- 10d ago First seen · 49 lines · 44 tokens per session scan A 403194461a63
ralph-mode is a skill published in the GitHub repository richardcb/oh-my-gemini (16 stars, last pushed 6mo ago), licensed MIT. It adds 44 tokens to every session and 362 once invoked, about $0.0002 per session on Opus 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 2026-08-30.
Other skills, from other repositories
intuitive-tests
Use this skill whenever the user asks about unit test best practices, test organization, flat test suites, redundant tests, test refactors, pytest/JUnit/Jest/xUnit layout, test taxonomy, flaky tests, coverage quality, fixtures, mocks, parametrization, pruning existing UTs, or "which tests are worth keeping." It…
intuitive-preflight
Turn a vague task, plan, issue, or "LGTM/go ahead" request into an approval-ready preflight contract before implementation starts. Use when the user wants prompt preflight, clearer scope, non-goals, context package, acceptance criteria, definition of done, verification, stop gates, the exact execution command, or…
intuitive-doc
Create and maintain an intuitive human documentation surface for AI-agent-developed repos. Use when humans should only need README.md, ARCHITECTURE.md, STATUS.md, and docs/human/ while planning logs, generated docs, retrospectives, ADR detail, and implementation evidence stay in AI-agent-only folders.
intuitive-refactor
Refactor and cleanup router for known code/module/API seams, stale surfaces, compatibility shims, architecture cleanup targets, changed-code quality review, oversized modules, repeated cleanup campaigns, and recurring whole-repo architecture maintenance goals. Use this when the user names a concrete seam, wants…
intuitive-shape
Shape a raw product or project idea into a bounded decision before planning or implementation. Use when deciding whether an idea deserves a bet, setting an appetite, comparing candidate bets under finite capacity, cutting scope, exposing rabbit holes and no-gos, or choosing BET, RESEARCH, RESHAPE, or PASS. This skill…
skill-runner
Run a bounded development task through named skills in an isolated, tmux-backed Codex or Claude session. Use for durable or artifact-sensitive worker phases, supervised skill runs, and post-run skill evaluation.