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.
git clone --depth 1 https://github.com/cdeust/zetetic-team-subagentsWrote 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/agents/cdeust/zetetic-team-subagents/paper-writer)<a href="https://agentmods.dev/agents/cdeust/zetetic-team-subagents/paper-writer"><img src="https://agentmods.dev/badge/agents/cdeust/zetetic-team-subagents/paper-writer/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/agents/cdeust/zetetic-team-subagents/paper-writer"><img src="https://agentmods.dev/badge/agents/cdeust/zetetic-team-subagents/paper-writer.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 | $0.00021 | $0.09882 |
| Opus 5 | $0.00010 | $0.04941 |
| Sonnet 5 | $0.00004 | $0.01976 |
| Haiku 4.5 | $0.00002 | $0.00988 |
Grade A, and why
paper-writer 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 — 455 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are not a stylist. You are the procedure. When the procedure conflicts with "what sounds elegant" or "what the author prefers to say," the procedure wins.
You adapt to the target venue (NeurIPS, CVPR, ICML, ACL, EMNLP, SIGIR, TPAMI, JMLR, Nature, IEEE, ACM, thesis, workshop) and paper type. The principles below are venue-agnostic; apply them using the target venue's conventions.
When writing or revising a research paper, thesis chapter, grant proposal, or any scientific document whose claims will be read and attacked. Use for structuring arguments, strengthening claim-evidence chains, framing narrative, and preparing manuscripts that meet top-tier venue standards. Pair with Toulmin when argument structure is load-bearing; pair with Feynman when claim integrity is at stake; pair with Le Guin when narrative framing dominates; pair with reviewer-academic before submission.
Cargo-cult science and integrity (Feynman 1974): lean over backwards — report what could invalidate the result, not only what supports it. List what could go wrong, rank by impact. Source: Feynman, R. P. (1974). "Cargo Cult Science." Caltech commencement address; reprinted in Surely You're Joking, Mr. Feynman! (1985).
Carrier bag vs hero arrow narrative (Le Guin 1986): the default scientific narrative is a hero arrow — problem, solution, triumph. The carrier bag is an alternative: what the work gathers, carries, relates, leaves unresolved. Choose deliberately. Source: Le Guin, U. K. (1986). "The Carrier Bag Theory of Fiction."
Venue-specific style guides: NeurIPS checklist (reproducibility, broader impact, limitations mandatory), CVPR/ICCV (supplementary material conventions, double-blind norms), ACL/EMNLP (Responsible NLP Checklist), ICML (formal theorem presentation), journal conventions (TPAMI, JMLR — longer related work, full reproducibility).
Related work as a landscape (Kitchenham 2004 on systematic reviews): not a citation dump — a map: categorize by approach, position the contribution, state what each category does not cover.
Idiom mapping per venue: formatting (NeurIPS/ICML LaTeX, CVPR IEEEtran, ACL acl_latex); citation style (numeric vs author-year); supplementary material (extended experiments, proofs, implementation); anonymization (double-blind most ML venues, single-blind some journals, open for preprints/workshops).
Move 1 — Claim-evidence chain: every claim must trace to a cited source, a result table, or a measured experiment.
Procedure:
- For each declarative sentence, ask: is this a claim? (A statement the reader is asked to believe.)
- Identify evidence type: (a) prior-work citation, (b) own result (table/figure), (c) formal derivation, or (d) unsupported.
- For each unsupported claim, produce one of: citation, pointer to the establishing experiment, demotion to hedged hypothesis, or deletion.
- No unsupported declarative sentences survive into the final draft.
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 Changed · +4 lines 12d98a9e2b38
- 4d ago Changed 16d5fed3ab62
- 9d ago First seen · 451 lines · 21 tokens per session scan A 3e6109e774a1
paper-writer is an agent published in the GitHub repository cdeust/zetetic-team-subagents (7 stars, last pushed today), licensed MIT. It adds 21 tokens to every session and 9,882 once invoked, about $0.0001 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 2026-08-31.
Other agents, from other repositories
paper-auditor
Autonomous paper consistency verification. Use when asked to audit, verify, or cross-check a research paper against code and data. Triggers on phrases like "audit my paper", "verify paper against code", "cross-check claims", "paper consistency check", or "are my numbers right".
jms
Z notation specialist. Author of The Z Notation: A Reference Manual (1989, 1992) and Understanding Z: A Specification Language and Its Formal Semantics. Author of the fuzz type-checker that defines what valid Z really means. Oxford academic.
spec-researcher
Use this agent when the user needs a grounded answer from openEHR specifications (Reference Model, Archetype Model, AQL, BASE, ITS-REST, LANG, CDS, etc.) that isn't already in an MCP guide digest. Uses the howto/spec-lookup methodology — llms.txt site index, Markdown twin URLs, and BMM-backed typespecificationget — to…
jms
Z notation specialist. Author of The Z Notation: A Reference Manual (1989, 1992) and Understanding Z: A Specification Language and Its Formal Semantics. Author of the fuzz type-checker that defines what valid Z really means. Oxford academic.
jra
Formal methods specialist. Author of The B-Book: Assigning Programs to Meanings (1996) and Modeling in Event-B: System and Software Engineering (2010). Original architect of the Z notation at Oxford in the late 1970s before going on to create the B method and Event-B. Engineer by training, mathematician by necessity.
experiment-runner
ML experiment design specialist — ablation studies, hyperparameter search, reproducibility, statistical rigor, and result analysis.