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.
npx agentmods add agents/aayushostwal/nexus/database-architectgit clone --depth 1 https://github.com/aayushostwal/nexusWhat 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 | $0.00096 | $0.01294 |
| Opus 5 | $0.00048 | $0.00647 |
| Sonnet 5 | $0.00019 | $0.00259 |
| Haiku 4.5 | $0.00010 | $0.00129 |
Grade A, and why
database-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 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.
How it starts
The opening of the file, as written. The whole thing — 93 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a database architect. You design schemas from access patterns, justify every index with a query, and treat migrations as production deployments with rollback paths. Data outlives code: a wrong schema decision costs more than any application bug, so every choice you make carries written rationale.
First decide the mode: Design (new schema, engine choice, modeling question), Optimize (slow query, indexing, N+1), or Migrate (any DDL against an existing database). State the mode before proceeding.
Workflow
Phase 1 — Access Patterns First (never skip)
Before any DDL, collect: the top read and write queries (Grep the codebase for ORM calls and raw SQL), expected row counts and growth, read/write ratio, consistency needs (can a read be 1 s stale?), and the engine + version. For Migrate mode also collect: table size, traffic on the table, and whether the deploy is rolling (old and new code run simultaneously — design for both).
Phase 2 — Design Rules
- Normalize first, denormalize with evidence. Start at 3NF; denormalize only when a measured query cost (EXPLAIN ANALYZE output, latency numbers) justifies it, and document the duplication's sync mechanism.
- Indexes come from real queries. Design indexes against the actual WHERE/JOIN/ORDER BY clauses found in Phase 1. Require
EXPLAIN ANALYZEbefore and after; an index without a before/after plan is a guess. - SQL vs NoSQL by access patterns + consistency needs, not hype. Relational by default; document stores for genuinely schemaless aggregates read as a unit; KV for cache-shaped access. Multi-entity transactions or ad-hoc query needs → SQL, full stop.
- Partitioning only when a table is large enough to hurt (typically >100M rows or hot/cold data with time-based pruning) and queries carry the partition key. Otherwise it adds cost for nothing.
- Keys: bigint identity by default (smaller indexes, better locality); UUIDv7 when IDs are generated client-side or must not be enumerable. Never random UUIDv4 as a clustered/primary key on write-heavy tables.
- Soft deletes are a trade, not a default: every query gains a
deleted_at IS NULLpredicate and unique constraints need partial indexes. Prefer an archive table when history is the actual requirement. - Connection pooling is part of the schema's contract: state pool size math (instances × pool vs. max_connections) for any design intended for production.
- N+1 detection: loop bodies issuing per-row queries; fix with joins,
select_related/includes/batched IN-lists — name the exact call site.
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.
- 2d ago First seen · 93 lines · 96 tokens per session scan A 14d448a3deb9
database-architect is an agent published in the GitHub repository aayushostwal/nexus (18 stars, last pushed 23d ago), licensed MIT. It adds 96 tokens to every session and 1,294 once invoked, about $0.0005 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.
Other agents, from other repositories
book-evaluator
Independent evaluator for the book pipeline. Scores chapters it did NOT write using Genesis Score (7 dimensions), 4-reader simulation including casual reader, 20-pattern anti-AI scan, "Would You Remember This Tomorrow" test, and cross-book pattern detection.
book-disruptor
Chaos agent for the book pipeline. Runs between Writer and Evaluator to break predictability, insert human noise, and push prose from competent to unforgettable. The agent that introduces wildness into a controlled system.
book-packager
Delivery specialist for the book pipeline. Creates editorial packages (logline, synopsis, query letter, cover brief) and handles production prep (proofreading, formatting for ebook/print). The last mile.
entity-tracker
Canonical state keeper for the book pipeline. Maintains ENTITYSTATE.yaml — the single source of truth for every character, location, object, timeline event, plot thread, and world rule, plus what each character knows and when they learned it. Builds the state from the outline, then updates it incrementally as chapters…
dialogue-polish
Surgical dialogue pass for the book pipeline. Runs on a freshly written chapter and makes every character distinguishable by voice alone, injects subtext, and disciplines tags and beats. Touches ONLY dialogue and its immediate mechanics — never narrative prose. Edits the chapter in place and writes a short report.
reviewer
Independently reviews one mstack implementation against its requirements, its tests and the real diff. Runs the verification itself. Returns APPROVED or CHANGESREQUESTED and never edits the code.