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/roomi-fields/paper-trailWrote 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/commands/roomi-fields/paper-trail/paper-trail-new-sota)<a href="https://agentmods.dev/commands/roomi-fields/paper-trail/paper-trail-new-sota"><img src="https://agentmods.dev/badge/commands/roomi-fields/paper-trail/paper-trail-new-sota/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/commands/roomi-fields/paper-trail/paper-trail-new-sota"><img src="https://agentmods.dev/badge/commands/roomi-fields/paper-trail/paper-trail-new-sota.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 | $0.00037 | $0.00840 |
| Opus 5 | $0.00018 | $0.00420 |
| Sonnet 5 | $0.00007 | $0.00168 |
| Haiku 4.5 | $0.00004 | $0.00084 |
Grade A, and why
paper-trail-new-sota 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 11d 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 — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/paper-trail:new-sota — Create a SOTA on a topic
Invoke the sota-writer skill to produce a new SOTA / literature
review on a given topic.
Usage
/paper-trail:new-sota <topic>
/paper-trail:new-sota <topic> --depth shallow|standard|exhaustive
/paper-trail:new-sota <topic> --max-candidates 30
What it does
- Pré-vol : invoke
python3 -m pipeline preflightto verify the environment (vault path, Python deps,paper-searchMCP registered). If pre-flight reportspaper-search MCP non enregistré, halt and show the install recipe printed by preflight before going further. The user must install the MCP, restart Claude Code, then re-run. - Phase A — Research : invokes
researchersub-agent for multi-source academic search (paper-search MCP across 22 platforms ; optionally NotebookLM + WebSearch) - Phase B — Acquisition : pushes candidates into the registry
with state
candidate, then invokespdf-cascadeto download + page 1 validation - Phase C — Reading : for each ref in
page1_validated, opens the PDF and writes structured notes in the ref's markdown body - Phase D — Writing : produces the SOTA at
adapter.sota_output_path(<topic_slug>), citing only refs inpage1_validated+, with a separate « Refs écartées » section listing rejected candidates and reasons
Safeguards
- Refuses to write from memory : every claim must trace back to
a
page1_validatedref's notes body - >30% DROP refusal : if more than 30% of candidates fail to
reach
page1_validated, refuses to write the SOTA and reports the situation (user must refine topic, enable shadow libs, or intervene manually on specific refs) - PreToolUse hook : the plugin blocks the final SOTA write if
any cited wikilink points to a non-
page1_validatedref (seetools/precheck_sota_wikilinks.py)
Shadow libraries
To maximize acquisition success, activate shadow libs before
running new-sota (see DISCLAIMER.md) :
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.
- 11d ago First seen · 95 lines · 37 tokens per session scan A 70447fff3a8a
paper-trail-new-sota is a command published in the GitHub repository roomi-fields/paper-trail (5 stars, last pushed 8d ago), licensed MIT. It adds 37 tokens to every session and 840 once invoked, about $0.0002 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 commands, from other repositories
lit-landscape
Generate figures visualizing literature trends, citation patterns, and research gaps.
verify-citations
Verify all citations against bibliography.
lit-review
A command for carrying out a systematic literature review, meaning a structured search and summary of research papers on a chosen topic. It also builds a citation graph showing connections between papers.
causal-design
Design or audit the identification strategy for an observational study. Use when the task concerns estimands, causal assumptions, threats to identification, or defensible research design rather than model implementation.
proof-readability
Improve the exposition and readability of a mathematical proof already verified as correct without changing its mathematics. Use when polishing a lemma, theorem, proof, or appendix after correctness checks. Not for verifying the proof; use $verify-math.
cite-check
Verify that citations actually exist and that the claims they support are faithful to the cited source. Runs deterministic existence checks (Crossref / OpenAlex / Semantic Scholar / arXiv) plus a claim-faithfulness pass via the alterlab-citation-verifier skill.