Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/WdBlink/autoresearch-papernpx agentmods add skills/wdblink/autoresearch-paper/mvp0Wrote 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/wdblink/autoresearch-paper/mvp0)<a href="https://agentmods.dev/skills/wdblink/autoresearch-paper/mvp0"><img src="https://agentmods.dev/badge/skills/wdblink/autoresearch-paper/mvp0/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/wdblink/autoresearch-paper/mvp0"><img src="https://agentmods.dev/badge/skills/wdblink/autoresearch-paper/mvp0.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 | $0.00110 | $0.02134 |
| Opus 5 | $0.00055 | $0.01067 |
| Sonnet 5 | $0.00022 | $0.00427 |
| Haiku 4.5 | $0.00011 | $0.00213 |
Grade A, and why
autoresearch-paper-mvp0 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 10d 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 — 192 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AutoResearch Paper MVP0
Use Codex as the Host and one physically separate Claude Code session with the frozen MiniMax model as the bounded Worker. P1–P5 remain the only scientific and experiment truth. P6 observes and advances those stores one transition per Codex heartbeat; it never creates a second research truth.
Required reading
Read these resources completely before acting:
mvp/README.md;mvp/prompts/codex-research-compiler.mdandmvp/schemas/research-ir.schema.jsonfor P1;- the Worker task/result schemas and
mvp/worker_adapter.pyfor P2; - the Experiment Receipt schema and
mvp/experiment_ledger.pyfor P3; - the evaluator/decision schemas and
mvp/evidence_gate.pyfor P4; mvp/prompts/codex-recompile-analyst.md, the P5 schemas, andmvp/recompile_loop.pyfor P5;mvp/prompts/codex-supervisor-heartbeat.md, the supervisor/runtime schemas,mvp/supervisory_controller.py, andmvp/runtime_assurance.pyfor P6.
Use examples/mvp0/fixed-wing-visual-guidance/ only as a shape example. Never
copy its claims, paths, evidence, or approval.
Non-negotiable boundaries
- P1 must discuss the research contract with the owner before freezing the
initial IR. Publish a proposal, stop for Human Critique, revise, stop for
Human Approval, then freeze
OWNER_REVIEWEDand replay it. ENGINEERING_ACCEPTANCEis test-only. Never use it for live work.- P6 may use
DELEGATED_ENGINEERING_REVIEWonly for a version N+1 delta that passes the deterministic execution-only policy and a fresh non-MiniMax Codex review. Problem, claim, baseline, metrics, thresholds, falsification, safety, evaluator semantics, and scientific hypothesis fields remain immutable. - The compiler, reviewer, revision author, and approver must be distinct
codex/<role>identities for delegated review. MiniMax is never a reviewer, approver, or final authority. - Never treat Worker
COMPLETED, P3 recording, or a Watchdog heartbeat as scientific success. Only the frozen P4 truth table can decide KEEP, PIVOT, STOP, or RECOMPILE. - Never repeat a failed task contract, reset its rejected worktree, rotate the
Worker model/session, or accept rejected files. A successor uses a clean new
worktree, an exact P5 freeze, the predecessor terminal receipt, the same
session UUID, and exact
--resume. - Unattended dispatch is forbidden until the full L0/L1/L2 activation receipt replays. A Codex automation file alone is not a complete Watchdog.
- One heartbeat performs at most one state transition. Stop after the committed tick and let the next heartbeat derive the next action.
- Pause on scientific/ambiguous changes, identity mismatch, unknown scheduler drift, exhausted deterministic recovery, or inconsistent lineage.
- Do not claim 24h, 7x24, production cutover, SOTA, or paper readiness from P6.
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.
- 10d ago First seen · 192 lines · 0 tokens per session scan A 4177de942aa6
autoresearch-paper-mvp0 is a skill published in the GitHub repository WdBlink/autoresearch-paper (2 stars, last pushed 29d ago), licensed MIT. It adds 110 tokens to every session and 2,134 once invoked, about $0.0006 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 skills, from other repositories
fba-simulator
Run Flux Balance Analysis (FBA) and related constraint-based simulations using COBRApy. Covers standard FBA, parsimonious FBA (pFBA), Flux Variability Analysis (FVA), loopless FBA, gene/reaction knockouts, and carbon source swapping. Outputs flux distributions and CSV files.
flux-analyzer
Analyse FBA flux distributions to extract biological insights. Covers gene essentiality, phenotypic phase planes, flux sampling, pathway-level aggregation, secretion product prediction, and production of publication- quality figures.
gsmm-validator
Validate a COBRApy genome-scale metabolic model for mass/charge balance, stoichiometric consistency, biomass producibility, dead-end metabolites, thermodynamic loops, and GPR rule formatting. Outputs a structured validation report with errors and warnings.
gsmm-builder
Build or load a genome-scale metabolic model (GSMM) using COBRApy. Covers loading from BIGG, constructing minimal models from scratch, setting medium constraints, and exporting validated .json model files.
stat-result-validator
Validate statistical research outputs for formulation quality, method-to- problem alignment, theory presence, experimental evidence, fair comparison, artifact completeness, and final-claim consistency.
statistical-experimental-evaluation
Design and run statistical experiments that test the formal problem, proposed methods, theoretical predictions, baselines, and ablations.