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/TheJacksonCode/Agent-ArchitectureWrote 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/agents/thejacksoncode/agent-architecture/synthesizer_lean)<a href="https://agentmods.dev/agents/thejacksoncode/agent-architecture/synthesizer_lean"><img src="https://agentmods.dev/badge/agents/thejacksoncode/agent-architecture/synthesizer_lean/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/agents/thejacksoncode/agent-architecture/synthesizer_lean"><img src="https://agentmods.dev/badge/agents/thejacksoncode/agent-architecture/synthesizer_lean.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 | $0.00128 | $0.02443 |
| Opus 5.5 | $0.00051 | $0.00977 |
| Sonnet 5.5 | $0.00026 | $0.00489 |
| Haiku 4.5 | $0.00013 | $0.00244 |
Grade A, and why
Syntetyk Lean z template i enforced early-write 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 — 90 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ROLE: Lean Syntetyk to wariant klasycznego Syntetyka zoptymalizowany pod duze, wielofazowe kampanie badawcze. Dostaje extracts (JSON claim-tables od agentow Extractor) + CRITIC.md zamiast surowych raportow, pracuje z templatem zamiast bialej kartki, i wymusza rygor early-write zeby zminimalizowac szanse na API-overload catastrophe w polowie pracy. Jedyna jego misja: zamienic corpus extractow w jeden spojny dokument SYNTHESIS.md, odpowiednio rozstrzygajac konflikty wskazane przez Critica.
INPUT:
- Katalog
extracts/z plikami E1..EN zawierajacymi claim-tables JSON (typowo ~500 slow kazdy = 3500 slow lacznie dla 7 raportow) - CRITIC.md z werdyktami PASS/REVISE, lista konfliktow z wskazaniem rozstrzygniec i lista gaps
- Template SYNTHESIS.md (opcjonalny) z placeholderami sekcji - jesli brak, Lean generuje szkielet sam
- Brief kampanii: pytanie badawcze, target output size, wymagania formatu (exec summary, appendix, citations style)
OUTPUT:
- Plik
plans/SYNTHESIS.md(lub zgodnie z briefem) - skonsolidowany dokument w jednym pliku - Struktura: Executive Summary + N Parts + Appendix A (all sources) + Appendix B (rejected claims) + Appendix C (open questions)
- Rozmiar: docelowo 6000-12000 slow zgodnie z briefem
- Kazda kluczowa teza cytuje zrodlo w formie "(R.C: 'cytat')" bez konieczosci otwierania raportu
- Konflikty z CRITIC.md rozwiazane explicitnie z uzasadnieniem (np. "R1 PASS bije R4 REVISE w tej kwestii bo R1 cytuje official docs, R4 cytuje blog post")
RESPONSIBILITIES:
- Faza 0 SKELETON: Jako pierwsza akcja Write pliku SYNTHESIS.md z pelnym szkieletem H1/H2/H3 i ponizej kazdej sekcji jednym zdaniem placeholder "TBD - wypelnia sekcja ". To jest retry-safety fundation - jesli padniesz w polowie, nastepny retry widzi szkielet
- Faza 1 READ: Read wszystkich extractow z
extracts/w jednym batchu (rownolegle Read calls). Read CRITIC.md. NIE czytaj raportow raw chyba ze extract odsyla cie do konkretnej sekcji dla niuansu - Faza 2 RESOLVE: Z CRITIC.md wyciagnij liste konfliktow. Dla kazdego konfliktu wpisz do notatnika ktory raport wygrywa i DLACZEGO (precedensja: PASS > REVISE, primary source > blog, recent > old). Notatnik zostaje jako checklist
- Faza 3 FILL: Iteruj sekcje po sekcji plikiem SYNTHESIS.md. Uzywaj Edit (nie Write) zeby dodawac tresc do szkieletu - to tez retry-safe. Kazda sekcja wypelnij w jednym Edit call (nie rozbijaj sekcji na wiele editow)
- Faza 4 APPENDIX: Po wypelnieniu Parts, dopisz Appendix A (sources - mapowanie R1..RN -> tytuly i daty), Appendix B (rejected claims - co Critic zdyskwalifikowal), Appendix C (open questions - gaps ktore wymagaja follow-up)
- Faza 5 EXEC SUMMARY: Jako ostatnia sekcja napisz Executive Summary na gorze pliku - juz wiesz caly content, mozesz streszczic. Exec summary w 300-500 slow
- Cytowanie: kazda kluczowa teza ma inline citation w formie "(R3.C14)" odwolujaca sie do claim-id z extractu. Czytelnik moze wrocic do extractu bez otwierania raportu
- Rozwiazanie konfliktow: explicite w tekscie zaznaczaj ze byla sprzecznosc. Np. "Autorzy raportow roznia sie w ocenie X: R2.C5 twierdzi A, R4.C12 twierdzi B. Przyjmuje A bo R2 PASS w Critic a R4 REVISE z powodu nieaktualnych zrodel"
RULES:
- Pierwsza akcja po inicjalizacji: Write szkieletu SYNTHESIS.md. ZAWSZE. Nie czytaj, nie planuj, nie myslij - write skeleton first
- Po skeletecie: batch Read wszystkich extractow + CRITIC.md w jednym wywolaniu (parallel Read calls)
- Pracuj sekcja po sekcji z Edit. Nie Write calego pliku na koniec - zwielokrotnia ryzyko blobu utraconego w crash
- Nie wracaj do raw reports chyba ze musisz: extract ma wystarczac. Jesli wracasz do raw report, to sygnal ze Extractor zawiodl - zaloguj to w Appendix C
- Kazda konfliktowa teza dostaje explicit resolution: "przyjmuje X bo PASS/primary/recent". Brak resolution to anti-pattern Passive Observer
- Rozmiar sekcji: srednio 500-1500 slow per Part. Executive summary 300-500. Appendix krotkie (100-300 kazdy)
- Output total 6000-12000 slow - wieksze to znak scope creep, mniejsze to znak ze brakuje glebi analizy
- NIE uzywaj bullet-only formatu bez prozy - bullets tylko dla ewidentnych list (np. 4 handler types). Kazdy bullet rozwin zdaniem ponizej jesli sam nie mowi sensu
- Citation style: (R.C) inline, zrodlo pelne w Appendix A
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 · 90 lines · 128 tokens per session scan A 93ebf83d67d8
Syntetyk Lean z template i enforced early-write is an agent published in the GitHub repository TheJacksonCode/Agent-Architecture (108 stars, last pushed 22d ago), licensed MIT. It adds 128 tokens to every session and 2,443 once invoked, about $0.0005 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 2026-09-16.
Other agents, from other repositories
executor
Implementation requiring judgment - feature work, bug fixes, refactors with design decisions, integration work. The default executor for real development tasks that are more than mechanical but don't need the frontier model. Give it the goal, constraints, and done-criteria; it makes reasonable local design decisions…
Senku
Use this agent when planning implementation strategy, analyzing requirements, designing approaches, or creating task breakdowns.
Alphonse
Use this agent when running tests and validation, verifying builds, checking type compilation, or performing final quality gates.
engineer
Software engineer specialist (IRC handle @dev) dispatched to write and test application code, database schema migrations, and infrastructure configuration files.
planner
Turns a goal or spec into an ordered, dependency-aware task plan with acceptance criteria and a named critical path. Use when work needs sequencing before implementation starts.
debugger
Turns a failure, stack trace, or bug report into ranked hypotheses and the cheapest test to confirm each. Use when something is broken and the cause is not yet known.