Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/jmagly/aiwgnpx agentmods add skills/jmagly/aiwg/ralph-externalWrote 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/jmagly/aiwg/ralph-external)<a href="https://agentmods.dev/skills/jmagly/aiwg/ralph-external"><img src="https://agentmods.dev/badge/skills/jmagly/aiwg/ralph-external.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 5.1 | $0.00018 | $0.02144 |
| Opus 5 | $0.00009 | $0.01072 |
| Sonnet 5 | $0.00004 | $0.00429 |
| Haiku 4.5 | $0.00002 | $0.00214 |
Grade A, and why
ralph-external 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 7d 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.
This is a copy
91% identical to agent-loop-ext — 33 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 242 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Skill access pattern (post-kernel-pivot, 2026.5+)
Skill names referenced in this document are AIWG skills, not slash commands. Most are not kernel-listed and cannot be invoked as
/skill-nameby the platform. Reach them via:aiwg discover "<capability>" aiwg show skill <name>Only kernel-listed skills (
aiwg-doctor,aiwg-refresh,aiwg-status,aiwg-help,use,steward) are directly invokable as slash commands. See skill-discovery rule.
Al External
You are the Al External Orchestrator — launching and managing crash-resilient iterative loops that run outside the AI session for long-running tasks.
Core Difference from ralph
ralph runs the loop inside the current AI session. ralph-external launches the loop as an external process via tools/ralph-external/run.sh, persisting all state to .aiwg/ralph-external/. If the session dies mid-loop, the loop survives and can be reattached or resumed.
Use ralph-external when:
- The task will take longer than a single session
- You need CI/CD pipeline integration
- You want crash recovery guarantees
- You need to run multiple loops in parallel
Natural Language Triggers
Users may say:
- "ralph external"
- "external ralph"
- "crash-resilient loop"
- "persistent ralph"
- "long-running ralph task"
- "ralph with crash recovery"
- "start background ralph"
Parameters
Objective (required)
The task the loop should accomplish. Passed as the first positional argument.
--completion (required)
Success criteria as a verifiable command. The loop exits when this command returns exit code 0.
Good examples:
--completion "npm test passes with 0 failures"--completion "npx tsc --noEmit exits with code 0"--completion "coverage report shows >80%"
--max-iterations (default: 10)
Maximum iterations before the loop halts and saves state for manual review.
--timeout (default: 60 minutes)
Maximum wall-clock time. Loop checkpoints state before exiting so it can be resumed.
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.
- 7d ago First seen · 242 lines · 18 tokens per session scan A d5fee83e409e
ralph-external is a skill published in the GitHub repository jmagly/aiwg (209 stars, last pushed today), licensed MIT. It adds 18 tokens to every session and 2,144 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. It is 91% identical to agent-loop-ext, differing in 33 lines, and is treated as a copy.
Other skills, from other repositories
goal-writer
Drafts a goal+rider document pair that briefs an autonomous coding agent on one round of work — a goal file under 4,000 characters (sized to fit the /goal command in both Claude Code and Codex) plus an unbounded rider with phased plans and named depth tests. Use when the user says "draft a goal", "write a goal+rider"…
surge
Use when a user provides a PRD, spec, or detailed requirements document and needs a full project delivered through iterative expert orchestration — multi-round analyze/research/design/implement/QA cycles with convergence detection. NOT for: single-file edits, quick prototypes, simple Q&A, or tasks without a written…
horizon
Run a durable Horizon workflow for a multi-feature goal with bounded autonomous retries and an audit trail.
check-in
Record a Parallax protocol checkpoint with concrete evidence before gated implementation work.
debug
Perform an evidence-based Parallax diagnosis or post-build audit and verify the repair.
hyperplan
Harden a non-trivial plan through a three-round adversarial critique and evidence-based synthesis.