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/alecs5am/ralphy/fixernpx skills add alecs5am/ralphy --skill fixergit clone --depth 1 https://github.com/alecs5am/ralphyWrote 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/alecs5am/ralphy/fixer)<a href="https://agentmods.dev/skills/alecs5am/ralphy/fixer"><img src="https://agentmods.dev/badge/skills/alecs5am/ralphy/fixer.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 | $0.00332 | $0.02270 |
| Opus 5 | $0.00166 | $0.01135 |
| Sonnet 5 | $0.00066 | $0.00454 |
| Haiku 4.5 | $0.00033 | $0.00227 |
Grade A, and why
fixer scanned grade A 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 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
2. **Apply fixes only through existing `ralphy` verbs.** No raw `ffmpeg` / `curl` / `bunx tsx`. If a fix needs a verb that doesn't exist, propose adding it and stop (AGENTS.md invariant #2). How it starts
The opening of the file, as written. The whole thing — 107 lines — stays where its author put it; the contents beside it link to each section on GitHub.
fixer — the eval-to-repair loop
The evaluator stops at the report (correct separation of concerns). You are the consumer: you turn that report into an executable, cost-gated fix pass and close the quality flywheel. The report is your input; a re-rendered, re-evaluated video with a better verdict is your output.
ALSO FIRE
- Immediately after
/evaluatorreturns awarnorfailverdict and the user reacts with "ok fix it" / "and now?" / "make it good" (any language — match their chat language, English on disk). - When the user drops an
eval.json/REPAIR_PLAN.mdpath and asks you to act on it. - When the agent-production contract's
repairphase (#406,ralphy project status <id> --contract) is the next gap and the user wants to proceed.
DO NOT FIRE
- For an UNRENDERED project, or one with no
eval.jsonyet — that is the editor (ralphy render) then/evaluatorfirst. You cannot repair what hasn't been evaluated. - For "rework scene 3 / rewrite the hook / shorten the VO" as scenario feedback on a draft — that is the scenarist playbook (
.agents/skills/scenarist/SKILL.md). The fixer is the eval-DRIVEN loop, not free-form script editing. - For false-positive triage only ("ignore the hook-static finding, it's expected for this template") — note the skip and move on; don't spin up a full repair pass.
HARD INVARIANTS
- No paid model call before approval. This is the load-bearing rule of the whole skill. The
repair-planverb is free and deterministic — run it freely. But the moment a fix is anart-directorre-roll or anyralphy generate .../ralphy render, you STOP and wait for the user's explicit "go" on the plan (or a prior "batch repair, don't ask me each time" opt-in). Every plan item is bornapprovalState: "pending"precisely so this gate is structural, not just prose. - Apply fixes only through existing
ralphyverbs. No rawffmpeg/curl/bunx tsx. If a fix needs a verb that doesn't exist, propose adding it and stop (AGENTS.md invariant #2). - Append-only. Re-rolls auto-version (
.v2); never--force-overwritewithout the user asking. The old eval, the old render, the old plan all stay on disk. - Deterministic plan, no LLM.
ralphy project repair-plandoes pure parsing of the eval output. Do not paraphrase or re-rank findings by hand — the verb owns the owner map + priority order.
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 · 107 lines · 332 tokens per session scan A 4fa53f0a3e53
fixer is a skill published in the GitHub repository alecs5am/ralphy (129 stars, last pushed 10d ago), licensed Apache-2.0. It adds 332 tokens to every session and 2,270 once invoked, about $0.0017 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
image-generation-gateways
Select, integrate, and operate multi-model image-generation gateways with model-specific schema discovery, version policy, asynchronous jobs, webhooks, spend approval, safe inputs and artifacts, data-governance review, and billing observability. Use for comparing or building against fal.ai, Replicate, or Together AI…
runway-video
Build and operate production-safe Runway API video generation, video editing, and character-performance workflows with native Runway models, explicit paid-call approval, duplicate-create protection, asynchronous task handling, secure media transfer, and evidence-aware governance.
kling-kolors-image
Plan, generate, edit, reference, and quality-control still images with Kling AI's hosted IMAGE surfaces or Kuaishou's open Kolors checkpoints. Use for Kling IMAGE 3.0/3.0 Omni/O1/2.1 web, official Kling CLI/MCP or Open Platform integration, and local Kolors text-to-image, image-to-image, IP-Adapter, ControlNet…
xai-grok-imagine-video
Produce, edit, extend, and govern short videos with xAI's direct Grok Imagine Video API. Use for text-to-video, image-to-video, multi-reference video, natural-language video edits, video continuation, exact media-cost approval, asynchronous request recovery, Files/Batch integration, moderation review, and API privacy…
product-ad-production
Diagnose, design, script, produce, adapt, and quality-control product video advertisements from evidence-backed briefs. Use for physical-product, software, app, service, ecommerce, direct-response, brand-response, demonstration, testimonial, launch, and variant ad work where product truth, legibility, claims, platform…
real-estate-content-production
Provider-independent real estate media production for listing photos, virtual staging, property videos, walkthrough reels, floor-plan explainers, neighborhood clips, renovation concepts, agent ads, and generated or retouched property visuals. Use when creating, editing, reviewing, prompting, packaging, or approving…