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/MadaraUchiha-314/the-loopWrote 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/madarauchiha-314/the-loop/do-task)<a href="https://agentmods.dev/commands/madarauchiha-314/the-loop/do-task"><img src="https://agentmods.dev/badge/commands/madarauchiha-314/the-loop/do-task/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/madarauchiha-314/the-loop/do-task"><img src="https://agentmods.dev/badge/commands/madarauchiha-314/the-loop/do-task.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 | $0.00048 | $0.01110 |
| Opus 5.5 | $0.00019 | $0.00444 |
| Sonnet 5.5 | $0.00010 | $0.00222 |
| Haiku 4.5 | $0.00005 | $0.00111 |
Grade A, and why
do-task 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 1mo 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 — 79 lines — stays where its author put it; the contents beside it link to each section on GitHub.
the-loop: do-task $ARGUMENTS
Do the thing the work item asks for, and nothing else. The process is the ad-hoc
loop (pdlc-adhoc-loop), the fourth shipped graph and the smallest one: three nodes,
no artifacts, no gates on files. Ask the-loop check <id> / the-loop graph show where
the item stands rather than re-deriving it. Load .the-loop/harness-config.yaml first
and honor every custom instruction doc it registers — that config is what tells you this
project's test, lint and type-check commands, and running them is the one piece of rigor
this loop keeps.
Before acting, read the the-loop skill — the rules that are not about phases still
apply in full: the paper trail, the self-authored marker on every comment you post
(reference/collaboration.md § loop prevention), context management, and the minimalism
ladder. Only the process is absent.
What this loop is
The requester armed this item with the-loop do instead of the-loop start. That is an
authorized human's explicit, recorded declaration that this work item runs without the
PDLC — it is frozen in work-item-state.json's loop field and the arming comment stands
on the thread. Honor it: the fastest correct path from the instruction to a working
change is the whole job.
work ⇄ review → complete
work— do it. Report back on the thread when you have something.review— the requester replies. Anything that is not "done" is more work, and routes straight back towork. Silence leaves the gate open.complete— the requester said it is finished. Closing the issue ends it too.
What is different from work-on and contribute-to
- The work item is the instruction. There is no
goal-definitiongate and noGoal:/Success criteria:format to wait for — unlikecontribute-to, which cannot start without one. Read the ticket title, body and thread; if the ask is genuinely ambiguous, ask on the thread and keep working on the unambiguous part, rather than stopping with nothing delivered.
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.
- 1mo ago Changed d7c6d9ddcd0d
- 1mo ago Changed 4f7c98650d6b
- 1mo ago First seen · 79 lines · 48 tokens per session scan A fe5fa3a3dd5c
do-task is a command published in the GitHub repository MadaraUchiha-314/the-loop (7 stars, last pushed yesterday), licensed MIT. It adds 48 tokens to every session and 1,110 once invoked, about $0.0002 per session on Opus 5.5. 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
prep
Prepare the meeting. One page from the record.
checklist-sop
Turn a process into a clear checklist or standard operating procedure (SOP) anyone can follow without you.
prioritize-tasks
Prioritize a task list by importance and urgency (Eisenhower matrix) and turn it into a clear order of action.
design-notification-system
Design notifications with a frequency budget, the right channel per message, and real user control.
improve
Make any piece of work better, keeping its intent and voice while fixing the weaknesses. A quick /improve command.
on
Turn on senior mode — refuses vague questions, hints instead of answers, never writes code.