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 Mathews-Tom/armory --skill research-critiquegit clone --depth 1 https://github.com/Mathews-Tom/armoryWrote 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/mathews-tom/armory/research-critique)<a href="https://agentmods.dev/skills/mathews-tom/armory/research-critique"><img src="https://agentmods.dev/badge/skills/mathews-tom/armory/research-critique/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/mathews-tom/armory/research-critique"><img src="https://agentmods.dev/badge/skills/mathews-tom/armory/research-critique.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to medium
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 →
- medium Prompt Injection · line 98 Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.Fix: Remove the large whitespace padding (blank-line blocks or long space runs) and review any content hidden below or to the right of it. Keep skill files compact and reviewable so no instructions can be
- medium Prompt Injection · line 104 Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.Fix: Remove the large whitespace padding (blank-line blocks or long space runs) and review any content hidden below or to the right of it. Keep skill files compact and reviewable so no instructions can be
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 | $0.00065 | $0.01837 |
| Opus 5 | $0.00032 | $0.00919 |
| Sonnet 5 | $0.00013 | $0.00367 |
| Haiku 4.5 | $0.00006 | $0.00184 |
Grade A, and why
research-critique 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 3d 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.
How it starts
The opening of the file, as written. The whole thing — 106 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Research Paper Critique
Purpose
Critically evaluate research papers as an engaged, intellectually honest colleague — not an adversarial reviewer generating decorative objections. The goal is understanding what a paper contributes, whether the evidence supports the claims, and where genuine tension exists between what is demonstrated and what is argued.
The failure mode you must avoid
The default behavior when critiquing a paper is adversarial credentialism: treating the paper as something to defeat, generating a volume of objections to demonstrate rigor, and mistaking the enumeration of limitations for insight. This produces critiques that are shallow, performative, and often incoherent — holding papers to standards that no published work in the field meets, demanding ablations that would constitute separate papers, and flagging limitations the author already disclosed as if discovering them.
The tell: if you retreat easily from a point when challenged, you were never committed to it. It was decorative, not substantive. Do not generate points you would abandon under pressure.
Analytical principles
1. Understand before you evaluate
Read the paper as the author intended it. What problem does it identify? What is the proposed mechanism or contribution? What is the experimental design trying to isolate? What claims does the paper actually make, versus what you assume it is claiming?
Most bad critiques attack a paper the author did not write. Get the real paper straight first.
2. Proportionality
Every empirical paper has limitations. Uncontrolled confounds, limited domains, incomplete ablations, finite sample sizes. These exist in every paper ever published, including the landmarks of the field. Listing them is not critique. The question is always: does this limitation actually undermine the paper's specific contribution?
A limitation that the author discloses, contextualizes, and accounts for in their claims is not a flaw you discovered. A limitation that would require a different experiment entirely to resolve is not a flaw — it is future work. A limitation that exists in every comparable study is not a differentiating weakness.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 3d ago First seen · 106 lines · 65 tokens per session scan A adc0207b41b8
research-critique is a skill published in the GitHub repository Mathews-Tom/armory (316 stars, last pushed 2d ago), licensed MIT. It adds 65 tokens to every session and 1,837 once invoked, about $0.0003 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 2026-09-05.
Other skills, from other repositories
or-solver
Unified solver detection, installation, selection, and license configuration for OR optimization skills (LP, MIP, SOCP). One-stop solver management: detect installed packages, install missing solvers with proper priority, select the best available solver by problem type, and troubleshoot common errors. Covers 20…
cdopt-optimization
Use when Codex needs to solve, reproduce, test, compare, generate examples for, or diagnose CDOpt manifold optimization workflows, including CDOpt package smoke tests, official problem description cards, Stiefel dictionary learning, SciPy wrappers, PyTorch/JAX/NumPy CDOpt examples, dependency checks, tiny CPU…
linear-programming
Models and solves linear programs (LP) with multiple solvers (COPT, Gurobi, MOSEK, CPLEX, HiGHS (scipy/highspy), CLARABEL, PuLP/CBC, CVXOPT, OR-Tools/GLOP, GLPK, SoPlex, lpsolve, ECOS). Restates natural-language word problems into LP before building and solving. When no local solver is available, searches GitHub for…
copt-linear-program
A tool for turning linear programming problems into mathematical models and solving them with Cardinal Optimizer, a software package for optimization. It can start from a word problem or from supplied matrices and JSON data.
osqp-solver
Model, solve, update, reproduce, and diagnose continuous convex quadratic programs with the OSQP Python interface, including canonical-form conversion, sparse CSC data, solver-status gates, residual checks, infeasibility certificates, warm starts, repeated parametric QPs, and academic citation. Use when a task names…
second-order-cone-programming
Models and solves second-order cone programs and problems convertible to SOCP form using cvxpy with multiple solver backends (COPT, Gurobi, MOSEK, CPLEX, CLARABEL, ECOS, SCS, CVXOPT, COSMO, OSQP). Handles linear objectives with second-order cone constraints, as well as problems reducible to SOCP such as portfolio…