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 Merck/ReactiveDynamics.jl --skill rd-framegit clone --depth 1 https://github.com/Merck/ReactiveDynamics.jlWrote 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/merck/reactivedynamics.jl/rd-frame)<a href="https://agentmods.dev/skills/merck/reactivedynamics.jl/rd-frame"><img src="https://agentmods.dev/badge/skills/merck/reactivedynamics.jl/rd-frame/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/merck/reactivedynamics.jl/rd-frame"><img src="https://agentmods.dev/badge/skills/merck/reactivedynamics.jl/rd-frame.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 | $0.00080 | $0.01426 |
| Opus 5.5 | $0.00032 | $0.00570 |
| Sonnet 5.5 | $0.00016 | $0.00285 |
| Haiku 4.5 | $0.00008 | $0.00143 |
Grade A, and why
rd-frame 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 today.
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.
How it starts
The opening of the file, as written. The whole thing — 68 lines — stays where its author put it; the contents beside it link to each section on GitHub.
rd-frame — from a question to a brief
Your output is BRIEF.md: a document in the operator's vocabulary where every slot nonetheless maps 1:1 onto the object model, so AUTHOR can compile it without a second interview. Read agent/studies/marginal_scientist/BRIEF.md first — it is what "already a model in all but syntax" looks like.
Interview, don't interrogate. Ask two or three questions at a time, propose a default for each, and let the user correct you. "I'll assume a trial holds one scientist for four months — is that right?" gets an answer; "how long does a trial take?" gets a meeting. Users know their system and do not know which details matter; deciding what matters is your job.
Start with the decision, not the system
The most common failure of this phase is modeling a system nobody has to decide anything about. So open with:
- What decision hangs on this? If nothing does, stop and ask what the user actually wants — a number nobody acts on is not worth a study.
- What is the lever? Something the user can actually change: a headcount, a budget, a priority rule, a deal, a policy. A study with no lever is a description, not a decision.
- What would make you act differently? This is the answer's shape — a threshold, a ranking, a go/no-go. Write it down before computing anything; it is what keeps you from producing a number the user cannot use.
Then run the qualifying test from rd-study explicitly and record the verdict in the brief. If it fails, say so and offer the arithmetic.
The slots
The brief's centerpiece is a table with one row per slot. Fill every row; an unfilled row becomes an invented default in phase 2, which is how models quietly answer the wrong question.
| Slot | What to ask | Where it lands |
|---|---|---|
| Scarce resource | "What runs out? What do people queue for?" | a plain place, init = how many |
| Reusable or spent? | "When work finishes, do you get it back?" — people/equipment yes, cash/materials no | conserved vs the default consumed arc modality |
| Metered resource | "Anything drawn continuously while work runs?" (budget, power) | a perstep arc out of a deep pool |
| Work in progress | "What flows through — and does each item have attributes that matter?" | a plain place if items are interchangeable; a structured place if each carries state (program, phase, value) |
| Starting state | "What is already in flight on day one?" | init for plain places, population[] for tokens |
| Intake | "Where does new work come from, and how fast?" | a transition, rate per tick |
| The process | "What does one unit of work consume, for how long, and how often does it succeed?" | a transition: lhs arcs, cycletime, prob_of_success |
| Contention rule | "When two things want the resource at once, who wins?" | priority weights |
| Outcome that carries value | "What produces the payoff, and when is it realized?" | a place with reward, produced on the rhs |
| Horizon and tick | "Over what period is this decided? What is the natural unit?" | meta.tspan, meta.dt |
| The lever | "What exactly changes between the arms?" | the diff between arm documents |
| Headline metric | "What number do you want?" | a named rdx metric |
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.
- today First seen · 68 lines · 80 tokens per session scan A 5fb326dfe352
rd-frame is a skill published in the GitHub repository Merck/ReactiveDynamics.jl (17 stars, last pushed 25d ago), licensed MIT. It adds 80 tokens to every session and 1,426 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-11-07.
rd-frame compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| rd-frameMerck/ReactiveDynamics.jl | Skill | 17 | 25d ago | MIT | A | 80 | Claude Code, Codex |
| julia-developsamtalki/AgentREPL.jl | Skill | 8 | 5mo ago | Apache-2.0 | A | 17 | Claude Code |
| julia-revisesamtalki/AgentREPL.jl | Skill | 8 | 5mo ago | Apache-2.0 | A | 20 | Claude Code |
| julia-evaluationsamtalki/AgentREPL.jl | Skill | 8 | 5mo ago | Apache-2.0 | A | 42 | Claude Code |
| julia-pkgsamtalki/AgentREPL.jl | Skill | 8 | 5mo ago | Apache-2.0 | A | 21 | Claude Code |
| julia-sessionsamtalki/AgentREPL.jl | Skill | 8 | 5mo ago | Apache-2.0 | A | 19 | Claude Code |
| julia-infosamtalki/AgentREPL.jl | Skill | 8 | 5mo ago | Apache-2.0 | A | 22 | Claude Code |