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/Khar-AG/t-800-agentnpx agentmods add commands/khar-ag/t-800-agent/t800-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/khar-ag/t-800-agent/t800-loop)<a href="https://agentmods.dev/commands/khar-ag/t-800-agent/t800-loop"><img src="https://agentmods.dev/badge/commands/khar-ag/t-800-agent/t800-loop/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/khar-ag/t-800-agent/t800-loop"><img src="https://agentmods.dev/badge/commands/khar-ag/t-800-agent/t800-loop.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 | $0.00000 | $0.01112 |
| Opus 5.5 | $0.00000 | $0.00445 |
| Sonnet 5.5 | $0.00000 | $0.00222 |
| Haiku 4.5 | $0.00000 | $0.00111 |
Grade A, and why
t800-loop 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 — 119 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/t800-loop — semi-manual Loop Engineering batch
Зачем: после прогона (report + lessons) — HITL-очередь в {memory_path}/loop-queue.md,
затем batch /t800-fix по fix-packs. Не автопродолжение через stop/followup.
Контракт: shared/loop-engineering-contract.md (v2).
Схема lessons: shared/lesson-schema-contract.md (v1.1 — status open|applied|rejected).
0. Discovery + pause
bash scripts/discover-target-project.sh --workspace "<WORKSPACE>" [--plugin-root "<PLUGIN_ROOT>"]
bash scripts/t800_loop_state.sh init --memory-path "<memory_path>"
Если существует {memory_path}/.loop-paused — стоп. Сообщи пользователю; не зови conductor.
1. Первая обязанность — conductor
Task(t-800-loop-conductor)
Передай полный контекст:
memory_pathplugin_rootrun_id(если известен; иначе conductor берёт последний run)- цель пользователя / HITL-заметки
Если Task(t-800-loop-conductor) недоступен:
Task(generalPurpose)
Промпт: полное содержимое agents/t-800-loop-conductor.md из плагина.
2. Материализация loop-queue
Conductor readonly — не пишет loop-queue.md. Родитель:
# stdin = JSON handoff (queue_patch) от conductor
python3 scripts/t800_loop_queue_write.py --memory-path "<memory_path>"
Путь: {memory_path}/loop-queue.md.
Open / Closed (Lesson Lifecycle v1.1):
- Open (
status=openили поле отсутствует) — кандидаты на HITL approve - Closed (
applied|rejected) — показать без action; не просить approve - Повторный
/t800-loopпри пустом Open → заглушка «нет открытых уроков», approve не просят
3. Опционально — classify / fixpack (только LOW + open)
Если lessons ещё без risk_class или нужен refresh:
python3 scripts/t800_risk_classifier.py --patch-file "<lesson.json>" --memory-path "<memory_path>"
Для LOW AND open only (после classifier, не LLM; schema v1.1):
python3 scripts/t800_lessons_to_fixpack.py --memory-path "<memory_path>" --lessons "<run_id|path/to/lessons.json>"
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 First seen · 119 lines · 0 tokens per session scan A f7cbaafef82d
t800-loop is a command published in the GitHub repository Khar-AG/t-800-agent (8 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,112 tokens. 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
quality-gate
Run formatter, linter, and type checks for a file or directory. Report remediation steps.
pn-audit-data
Design or review database schema — normalization, indexing strategy, migration plan, and data model decisions. Surgical command for database design. Use standalone or as part of pn-backend-audit.
pn-audit-performance
Identify and fix backend performance issues — N+1 queries, missing cache layers, blocking I/O, slow queries, and connection pool exhaustion. Surgical backend performance pass.
pn-delight
Add purposeful moments of joy, motion, and personality. Use when the design is correct and complete but feels flat and lifeless.
create-repro
Isolate a bug report into a minimal, standalone reproduction script or test case.
cleanup
Run a systematic cleanup on the provided code to resolve smells, simplify complexity, and enforce best practices.