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 skills/dgk-dev/dgk-gpt/ralphnpx skills add dgk-dev/dgk-gpt --skill ralphgit clone --depth 1 https://github.com/dgk-dev/dgk-gptWrote 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/dgk-dev/dgk-gpt/ralph)<a href="https://agentmods.dev/skills/dgk-dev/dgk-gpt/ralph"><img src="https://agentmods.dev/badge/skills/dgk-dev/dgk-gpt/ralph.svg" alt="Measured on agentmods" 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 | $0.00046 | $0.00269 |
| Opus 5 | $0.00023 | $0.00134 |
| Sonnet 5 | $0.00009 | $0.00054 |
| Haiku 4.5 | $0.00005 | $0.00027 |
Grade A, and why
ralph 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 5d 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
Use this as a persistence override, not a separate methodology.
Default
- Treat the task as incomplete until the requested outcome is verified.
- Repeat the loop: inspect -> implement -> verify -> fix -> verify again.
- If a check fails, continue with the next fix instead of stopping at the first attempt.
- If one approach stalls, switch approaches and keep going.
- Keep intermediate updates short and progress-focused.
Do Not Stop For
- "Probably fixed"
- "Good enough"
- Partial completion
- A failed first or second attempt
- Needing one more verification pass
Stop Only When
- The requested outcome is verified.
- A real external blocker prevents further progress.
- The user explicitly stops or redirects the work.
Avoid
- Do not ask whether to continue after each failure.
- Do not declare success without running the relevant checks.
- Do not fall back to planning mode unless execution is genuinely blocked.
Finish
Return:
- what was completed
- what was verified
- what is still blocked, if anything
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.
- 5d ago First seen · 44 lines · 46 tokens per session scan A b56eeedcc652
ralph is a skill published in the GitHub repository dgk-dev/dgk-gpt (53 stars, last pushed 5mo ago), licensed MIT. It adds 46 tokens to every session and 269 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
cancel
Cancel an active tracked Claude Code job in this repository. Args: [job-id]. Use only when the user wants to stop a queued or running Claude Code job.
eco-max
Maximum-savings variant of eco - the same frugality rules with a tighter reply budget, for routine chores (rename, small fix, quick lookup, boilerplate). Prefer plain eco for hard or high-stakes work. Works in any language.
next
Recommend the next one to three concrete actions with the highest leverage for the latest actionable result or current Binding. Use only when the user invokes next or asks what to do next; never redirect or act silently.
eco
Token-frugal mode for Codex - fewer tokens per turn at full task quality. Use when the user mentions tokens, cost, budget, quota, rate limits, or working economically, in any language. Say nothing about invoking it - read this file, then let your first visible action be the work. Pass "setup" to configure durable…
codex-conductor
Situational awareness across local Codex sessions. Use when the user says "sort out my Codex sessions", "what is Codex doing", "summarize my Codex windows", "codex conductor", or "what's running".
eco-long-body
Token-frugal mode for Codex - minimize token spend (no preamble turns, no unasked surveys, region reads, batched shell calls) at full task quality. Use when the user mentions tokens, cost, budget, quota, rate limits, or economical operation, in any language. Invoke once; applies for the rest of the thread. Pass…