taupr

An automated work loop that handles one pull request task per round. A pull request is a proposed code change submitted for review; the loop prioritizes maintenance before starting new work.

In plain words
What is it for?
Use it to rebase pull requests, fix continuous-integration failures, address review findings, review stalled pull requests, or open a new pull request when capacity allows.
Why use it?
It prevents new pull requests from piling up while existing ones have merge conflicts, failed builds, or review findings. It also limits how many of its own pull requests can be open at once.

Command

Install

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.

agentmods
npx agentmods add commands/cbirkbeck/mathlib-quality/taupr
Clone the repo
git clone --depth 1 https://github.com/CBirkbeck/mathlib-quality
Per session 78 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,353 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00078 $0.04353
Opus 5 $0.00039 $0.02176
Sonnet 5 $0.00016 $0.00871
Haiku 4.5 $0.00008 $0.00435

Measured 2d ago against content hash be725e683e42, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

taupr 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 2d 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.

commands/taupr.md · 369 lines

How it starts

The opening of the file, as written. The whole thing — 369 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/taupr — the Tau Ceti worker loop

Modelled on the reference worker (kim-em/TauCetiWorker): a round does exactly one unit of work — the first of these that applies.

R1  REBASE     one of our PRs has a genuine TauCeti/ conflict after a sibling merged
R2  FIX-CI     one of our PRs has `build` red — it cannot be reviewed until it builds
R3  FIX        one of our PRs has findings: fix the code, or contest a wrong one in-thread
R4  REVIEW     one of our PRs is green but CI has not reviewed it for ≥ 1h  →  --post
R5  AUTHOR     otherwise: open a new PR advancing a target — ONLY if fewer than
               --max-open (default 3) of our PRs are already open, else IDLE

Maintenance outranks authoring. R1–R4 come first so PRs already in flight cannot be starved by opening yet more of them. A round that finds work at R2 stops there.

R5 is capped at 3 of our own PRs open. This matters more than it looks. A PR that is still building is not red, has no findings, and is not yet reviewable — so no step catches it, the board reads as quiet, and an uncapped round falls through to authoring. On a ten-minute loop that is a new PR every tick. Worse, R4's one-hour wait means a green PR looks like nothing to do for a whole hour, which is exactly when the loop would be busiest authoring.

So: if 3 or more of our PRs are open, R5 does not run and the round reports IDLE. Idle is a correct outcome — it means the work in flight is waiting on someone else. Change it with --max-open <n>; --max-open 0 lifts the cap.

Count PRs we authored that are still open, in any state. Merged and closed do not count, so the cap self-releases as work lands.

CI reviews our PRs; we do not race it. Opening the PR is enough — pr-build runs, and a green build triggers the review automatically. Self-reviewing costs your subscription and your wall-clock, so R4 exists only as a fallback for PRs CI has left sitting.

Merging, closing and de-duplicating are the repo's CI, not this loop. Do not merge green PRs, close stuck ones, or sweep duplicates from here.

A GitHub API failure aborts the round. It must never read as "nothing to do" and fall through to R5 — a transient outage would then author duplicate PRs. No data means stop, not proceed.

Read the full file on GitHub · 369 lines

Changes

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.

  1. 2d ago First seen · 369 lines · 0 tokens per session scan A be725e683e42

Subscribe to this mod's changes

taupr is a command published in the GitHub repository CBirkbeck/mathlib-quality (32 stars, last pushed 13d ago), licensed MIT. It adds 78 tokens to every session and 4,353 once invoked, about $0.0004 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 2026-08-30.