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 instructions/nolainjin/paper-verify/agents-mdgit clone --depth 1 https://github.com/nolainjin/paper-verifyWrote 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/instructions/nolainjin/paper-verify/agents-md)<a href="https://agentmods.dev/instructions/nolainjin/paper-verify/agents-md"><img src="https://agentmods.dev/badge/instructions/nolainjin/paper-verify/agents-md.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 5.1 | $0.00588 | $0.00588 |
| Opus 5 | $0.00294 | $0.00294 |
| Sonnet 5 | $0.00118 | $0.00118 |
| Haiku 4.5 | $0.00059 | $0.00059 |
Grade A, and why
paper-verify AGENTS.md 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 6d 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 — 47 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — paper-verify
Short entrypoint for coding agents. Detail lives in linked files (progressive disclosure).
What this is
A citation-verification tool (extract → fetch → judge → score → report). Pure-Python,
local-first. See README.md for the product; pyproject.toml for deps.
Feature list is the source of truth (harness primitive)
feature_list.json is the single record of what must work and whether it does — not a
human note. Each feature = (behavior, verification command, state, evidence).
Rules (do not bypass):
- One feature
activeat a time. Pick the nextnot-startedwhosedependenciesarepassing. - A feature reaches
passingONLY via its verification command exiting 0. Never hand-editstatetopassing. The gate controls the transition:source .venv/bin/activate python tools/feature_gate.py # report drift / LIEs (state=passing must verify) python tools/feature_gate.py --sync # promote verified features to passing (+evidence) state: passingwith a failing verification is a GATE FAIL (exit 1) — the false-"done" guard. Wirefeature_gate.pyinto pre-commit / CI to make it un-bypassable.- Back-pressure = count of non-passing features. 0 = done.
Verify before claiming done
- Full suite:
python -m pytest -q(must stay green). - The feature gate is the per-feature gate; the suite is the global gate.
Dev setup
- Recreate the venv if missing:
python -m venv .venv && .venv/bin/pip install -e '.[dev]' - Full suite:
.venv/bin/python -m pytest -q
Provider neutrality (hard constraints)
- Do not rewrite the verifier core for one provider; no permanent provider
branches — provider differences live in
paperverify/harness/anddocs/providers/. - Keep tests network-free (monkeypatch fetch/judges) unless clearly marked as integration.
- Generated
*_report.md/*_claims.jsonlstay out of commits (.gitignorecovers them).
Conventions
- SSoT: schema version in
paperverify/report.py(SCHEMA_VERSION); IPC/shape changes bump it. - Backlog/audit:
docs/harness/2026-06-04_paper-verify-audit/(note: a static doc — it drifts;feature_list.json+ the gate are authoritative for current state).
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.
- 6d ago First seen · 47 lines · 588 tokens per session scan A 66c6f35756a0
paper-verify AGENTS.md is an instructions file published in the GitHub repository nolainjin/paper-verify (5 stars, last pushed 23d ago), licensed MIT. It adds 588 tokens to every session, about $0.0029 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 instructions, from other repositories
paperbanana copilot-instructions.md
Instructions for llmsresearch/paperbanana, covering copilot instructions for paperbanana, build & test, install for development, run full test suite and run a single test file.
daily_arxiv_digest AGENTS.md
Instructions for yang3kc/daily_arxiv_digest, covering agents.md, project overview, commands, development and running and testing.
research-claw CLAUDE.md
Instructions for nanoAgentTeam/research-claw, covering claude.md, 开发指令, 核心执行, 环境配置 and 代码架构.
machine-learning-library AGENTS.md
AGENTS.md instructions for ATOM00blue/machine-learning-library, covering agent guide — machine-learning-library, layout, experimental aprl algorithm, frontmatter schema and retrieval workflow (do this).
machine-learning-library CLAUDE.md
Claude Code instructions for ATOM00blue/machine-learning-library, covering claude.md and quick retrieval protocol.
EFC-Plugin copilot-instructions.md
Copilot instructions for Nlai741533/EFC-Plugin, covering copilot coding agent instructions, project overview, critical rules, security — never commit secrets and test-first development.