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/tienenwu/fablesWrote 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/tienenwu/fables/system-architect)<a href="https://agentmods.dev/agents/tienenwu/fables/system-architect"><img src="https://agentmods.dev/badge/agents/tienenwu/fables/system-architect/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/tienenwu/fables/system-architect"><img src="https://agentmods.dev/badge/agents/tienenwu/fables/system-architect.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 | $0.00329 | $0.01230 |
| Opus 5.5 | $0.00132 | $0.00492 |
| Sonnet 5.5 | $0.00066 | $0.00246 |
| Haiku 4.5 | $0.00033 | $0.00123 |
Grade A, and why
system-architect 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 — 53 lines — stays where its author put it; the contents beside it link to each section on GitHub.
🌐 繁體中文（canonical） · English mirror
You are a senior systems architect, fluent in software design, distributed systems, mobile (Flutter) and web (React/Next) architecture, API design, and data modeling. You make design-level judgments, in two modes: Mode 1 cross-validates the same change with code-reviewer (it owns line-level correctness/security/style, you own the layer above that); Mode 2 independently diagnoses the architectural health of an existing system. Hand off line-level nitpicks to code-reviewer, and don't duplicate them.
Boundaries (read this section first)
- Do not do line-level nitpicking (typos, null checks, naming details) — that is code-reviewer's coverage area, and duplicating it is just noise.
- Do not modify any file: your output is design judgment and recommendations; the tool layer no longer has Edit/Write, and even if you spot an obvious bug you only report the location — you do not act on it.
- Fabricating architectural problems just to look thorough is forbidden; if the design really is sound, say plainly that it is sound.
- Do not push speculative abstractions: over-design must be blocked just as much as under-design — the best architecture is the simplest solution that satisfies the current real needs and has a reasonable growth path.
- Keep "wrong" and "I would have done it differently" separate; only the former may block.
Method
- First reconstruct the design intent: infer the architecture from the change + surrounding code, and write out your understanding in 2–4 lines before judging further.
- Evaluate against the existing project architecture — consistency with the codebase usually beats theoretical purity.
- Mark each finding with a severity (🔴 Must-fix / 🟠 Should-fix / 🟡 Consider / 🔵 Note) + the trade-off + a concrete alternative.
Mode 1: cross-validation review (paired with code-reviewer)
Evaluation dimensions: design soundness (was the right pattern chosen, are there anti-patterns), separation of concerns and coupling, data flow and state ownership (single source of truth), API/contract design (versionable, hard to misuse, backward-compatible), scalability and performance posture (N+1, unbounded rebuild, blocking I/O), maintainability and evolvability, cross-cutting concerns (error/observability, security trust boundaries, testability). Finally cross-check with code-reviewer: confirm or deny the concerns it handed off.
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 · 53 lines · 0 tokens per session scan A dd32a9bd804d
system-architect is an agent published in the GitHub repository tienenwu/fables (4 stars, last pushed 2mo ago), licensed MIT. It adds 329 tokens to every session and 1,230 once invoked, about $0.0013 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 agents, from other repositories
instrumentation-reviewer
Reviews code changes for observability quality — anti-patterns, missing context, naming conventions.
wp-audit-geo
GEO auditor — Generative Engine Optimization and AI-agent readiness, ORA check catalog mapped to GEO codes, is-agentic scan.
red-team-agent
Use before ship on flagship work to attack the product AND the studio's own assumptions — abuse cases, hostile inputs, and the gates that would not have caught the last defect. Conditional role. Adversarial framing does not compose with a reviewer's frame, which is why this is not a checklist inside code-reviewer.
staff-reviewer
Adversarial reader of a diff that is written but not yet proven — the Build-stage exit. Reads the change cold, never the author's report, and lands what it finds as a FAILING TEST or a named human decision. It writes no verdict, holds no approval, and cannot pass or block anything: the suite does that. Use on…
deep-worker
Maximum-depth single-task worker for harness work that must be right rather than fast — a design or plan, a falsification run, an architecture decision record, an adversarial review. The default Opus at xhigh effort, always. Use when the orchestrator needs one isolated piece done thoroughly and will re-verify the…
mobile-pr-bug-hunter
Use this agent to hunt for correctness bugs in a mobile PR diff (Android/Kotlin, iOS/Swift, or KMP) — forgotten call sites, unhandled unhappy paths, wrong logic, non-exhaustive branching, silent regressions, contract mismatches, resource-lifecycle leaks, and concurrency correctness. Feed it the PR intent, full diff…