ralphex is a command-line tool that runs coding agents through implementation plans, executing repository tasks autonomously in separate sessions. It is intended for developers who want long-running features implemented, validated, reviewed, and committed with less interactive supervision.
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 umputun/ralphex --skill ralphex-plangit clone --depth 1 https://github.com/umputun/ralphexWrote 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/umputun/ralphex/ralphex-plan)<a href="https://agentmods.dev/skills/umputun/ralphex/ralphex-plan"><img src="https://agentmods.dev/badge/skills/umputun/ralphex/ralphex-plan/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/umputun/ralphex/ralphex-plan"><img src="https://agentmods.dev/badge/skills/umputun/ralphex/ralphex-plan.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Prompt Injection · line 182 Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.Fix: Audit all comments and invisible characters. Remove any instructions that direct the agent to perform unauthorized actions. Use plain, reviewable content.
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 | $0.00009 | $0.02831 |
| Opus 5.5 | $0.00004 | $0.01132 |
| Sonnet 5.5 | $0.00002 | $0.00566 |
| Haiku 4.5 | $0.00001 | $0.00283 |
Grade C, and why
ralphex-plan scanned grade C with 1 finding 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 1mo 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.
Hidden instructionshighPrompt injection
Directives inside HTML comments, invisible characters or bidirectional overrides are read by the model and not by the person reviewing the file.
<!-- Task structure guidelines: - Each task = ONE logical unit (one function, one endpoint, one component) - Use specific descriptive names, not generic "[Core Logic]" or "[Implementation]" - Aim for ~5 checkboxes per ta How it starts
The opening of the file, as written. The whole thing — 315 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Implementation Plan Creation
Create an implementation plan in docs/plans/YYYYMMDD-<task-name>.md with interactive context gathering.
Prerequisites: Verify CLI Installation
Check if ralphex CLI is installed (needed to execute the plan after creation):
which ralphex
If not found, inform user they'll need it to execute the plan:
- macOS (Homebrew):
brew install umputun/apps/ralphex - Linux (Debian/Ubuntu): download
.debfrom https://github.com/umputun/ralphex/releases - Linux (RHEL/Fedora): download
.rpmfrom https://github.com/umputun/ralphex/releases - Any platform with Go:
go install github.com/umputun/ralphex/cmd/ralphex@latest
Proceed with plan creation regardless, but remind user to install before execution.
Step 0: Parse Intent and Gather Context
Before asking questions, understand what the user is working on:
-
Parse user's command arguments to identify intent:
- "add feature Z" / "implement W" → feature development
- "fix bug" / "debug issue" → bug fix plan
- "refactor X" / "improve Y" → refactoring plan
- "migrate to Z" / "upgrade W" → migration plan
- generic request → explore current work
-
Launch Explore agent (via Task tool with
subagent_type: Explore) to gather relevant context based on intent:For feature development:
- locate related existing code and patterns
- check project structure (README, config files, existing similar implementations)
- identify affected components and dependencies
For bug fixing:
- look for error logs, test failures, or stack traces
- find related code that might be involved
- check recent git changes in problem areas
For refactoring/migration:
- identify all files/components affected
- check test coverage of affected areas
- find dependencies and integration points
For generic/unclear requests:
- check
git statusand recent file activity - examine current working directory structure
- identify primary language/framework from file extensions and config files
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.
- 1mo ago First seen · 315 lines · 9 tokens per session scan C 683b69b3bd33
ralphex-plan is a skill published in the GitHub repository umputun/ralphex (1,504 stars, last pushed 3d ago), licensed MIT. It adds 9 tokens to every session and 2,831 once invoked, about $0.0000 per session on Opus 5.5. A static security scan graded it C with 1 finding (hidden instructions). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
claude-loop
Autonomous orchestrated loop that works through a prd.json story-by-story using a fresh Task subagent per story. Use this skill whenever the user asks to "run the loop", "claude loop", "start the loop", "kick off the loop", "work through the prd", "process prd.json", "implement all the stories", "iterate through the…
agent-loop
Run bounded, resumable OpenAI-native coding loops with shell and applypatch, similar to Claude Code agent loops. Use when the user asks for a loop, autonomous iteration, or a persistent inspect-edit-test cycle with a turn cap.
humanizer
Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague…
html-ppt-zhangzara-vellum
A humanities lecture: how Renaissance linear perspective reshaped early cartography — sources, argument, and evidence. Built as a decision-grade academic research deck for faculty, graduate seminar.
code-review
Review code for quality, maintainability, and correctness.
rubber-duck
Adversarial "rubber duck" review that turns explaining-out-loud into a hallucination check. The main session is the PRESENTER (it did the work — a design doc, investigation, or analysis — and holds the real reasoning) and reconstructs the topic to a LISTENER — a spawned subagent pinned to a DIFFERENT-vendor model that…