Borrowing it
Nothing to install: this file belongs to dannyota/banhmi. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/dannyota/banhmi/master/AGENTS.mdgit clone --depth 1 https://github.com/dannyota/banhmiWrote 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/instructions/dannyota/banhmi/agents-md)<a href="https://agentmods.dev/instructions/dannyota/banhmi/agents-md"><img src="https://agentmods.dev/badge/instructions/dannyota/banhmi/agents-md/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/instructions/dannyota/banhmi/agents-md"><img src="https://agentmods.dev/badge/instructions/dannyota/banhmi/agents-md.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 | $0.03090 | $0.03090 |
| Opus 5.5 | $0.01236 | $0.01236 |
| Sonnet 5.5 | $0.00618 | $0.00618 |
| Haiku 4.5 | $0.00309 | $0.00309 |
Grade A, and why
banhmi AGENTS.md 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 4d 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 — 188 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Repository rules
Project status
Development and maintenance have stopped. Retain the repository as an archive. Hosted endpoints are unsupported; deployment descriptions are historical. This status supersedes earlier delivery, maintenance, and cutover directions in the retained plans and designs. Do not resume development without new owner instructions. Repository safety, review, and Git rules still apply to authorized archival work.
Authority and reading order
AGENTS.md is the single current rule authority. Topic files add task-specific safeguards.
Direct owner instructions take precedence. If project documents conflict, stop dependent work and
resolve the shared documentation with the change.
Read the documentation index, architecture, the applicable
topic rules below, and the spec or design for the task. Read only the delivery plan
sections the task touches. Designs are proposals unless labeled approved or implemented.
Keep durable decisions in tracked repository records, not tool memory.
Read short-term work for the current execution queue. The owning plans retain scope,
milestone acceptance, evidence, and open decisions.
Product and evidence boundaries
- The former planned product was Tra cứu quy định at
tracuuquydinh.vn, the approved, unbuilt direction in Website. banhmi remains the engine, repository, and corpus name. - The archived scope prioritized Vietnam and retained Malaysia, Singapore, Indonesia, Thailand, and
Cambodia on their existing
danny.vnhostnames. That maintenance plan is no longer active. - VNGCloud in Vietnam was the approved, unbuilt target for the new Vietnam system: serving, corpus writer and workers, review services, and artifact and backup storage. Keep writer and worker capacity separate from the serving host in the archived design. The Vietnam cutover is no longer planned. The former hosting requirements remain in the hosting decision.
- Serve evidence only. Our servers run no answer model, generative model, or reranker model.
The query embedder is the only served model and runs on our own host. The user's own AI reads
and judges evidence through the ChatGPT plugin or through
/mcpin Claude and other clients. - Decision: Reject a platform-run model and a reranker. Why: Sending user questions abroad is a cross-border transfer under Luật Bảo vệ dữ liệu cá nhân 91/2025/QH15, Điều 20. A reranker adds cost and latency with no gain over the user's own model. Rejected: Third-party question processing and query-time reranking. Detail: Public legal source documents may use approved cloud batch OCR and bulk embedding.
- Vietnam search accepts Vietnamese text with diacritics or a document number such as
50/2024/TT-NHNN. English is for interface text only. This requirement is approved, not deployed; the deployed MCP brief still says to translate first. - Keep five MCP tools:
guide,corpus_status,quality_gaps,search, anddocument. - Keep a separate corpus and database per jurisdiction on one shared codebase. Preserve each jurisdiction's sources, language, parser, citations, validity, and evaluations. Evidence retains exact native citations, validity, relations, provenance, confidence, and known gaps. Accuracy takes priority over breadth. Never hide weak or missing evidence.
- Legacy and new corpus code are separate. Legacy is the v1 corpus code that serves the five
maintained countries and the current Vietnam service:
pkg/{app,ingest,extract,pipeline},pkg/base/{jurisdiction,sourceurl}, the migration chains underdeploy/migrations/, the query sources undersql/{bronze,silver,gold,ingest,config}, and their generated stores underpkg/store/{bronze,silver,gold,ingest,config}. Auth code, its chain, andpkg/store/authare not legacy. Legacy is frozen: no new migrations, no new features, and no behavior change without an owner-approved maintenance fix. Legacy tests stay and run until the slice that retires their package deletes both together. New Vietnam code follows the approved layout. - Decision: Build the new Vietnam corpus schema on fresh migration chains and freeze the legacy
chains.
Why: The Vietnam writer is a new database with no data to migrate. Appending to the legacy
chains would carry their constraints, unused tables, and compatibility columns into a database
that never held them.
Rejected: Additive migrations on the legacy five-schema chains and a nullable legacy
build_id. Detail: The corpus database spec names the chains, query sources, and generated-store paths. The engineering rules govern chain mutability and hosted generation. - Report implemented code as coded. Report corpus behavior as validated only after checks against real documents. Report deployed behavior only after checking the deployed service.
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.
- 4d ago Changed · +8 lines · +74 tokens per session d77953178dae
- 24d ago Changed · +30 lines · +581 tokens per session 5e155ea85f0d
- 25d ago Changed · +2 lines · +34 tokens per session d56df6331108
- 26d ago First seen · 148 lines · 2,401 tokens per session scan A 29ff5c25f145
banhmi AGENTS.md is an instructions file published in the GitHub repository dannyota/banhmi (2 stars, last pushed 6d ago), licensed Apache-2.0. It adds 3,090 tokens to every session, about $0.0124 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 2026-10-12.
banhmi AGENTS.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| banhmi AGENTS.mddannyota/banhmi | Instructions file | 2 | 6d ago | Apache-2.0 | A | 3,090 | Codex, OpenCode |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 143k | today | MIT | A | 7,461 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 5,516 | GitHub Copilot |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | today | MIT | A | 4,680 | Codex, OpenCode |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 3,093 | GitHub Copilot |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 275k | 3d ago | MIT | A | 4,313 | Codex, OpenCode |