Borrowing it
Nothing to install: this file belongs to ROCm/aiter. 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/ROCm/aiter/main/.github/instructions/aiter-ops-flydsl.instructions.mdgit clone --depth 1 https://github.com/ROCm/aiterWrote 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/rocm/aiter/aiter-ops-flydsl)<a href="https://agentmods.dev/instructions/rocm/aiter/aiter-ops-flydsl"><img src="https://agentmods.dev/badge/instructions/rocm/aiter/aiter-ops-flydsl/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/rocm/aiter/aiter-ops-flydsl"><img src="https://agentmods.dev/badge/instructions/rocm/aiter/aiter-ops-flydsl.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 | $0.00785 | $0.00785 |
| Opus 5.5 | $0.00314 | $0.00314 |
| Sonnet 5.5 | $0.00157 | $0.00157 |
| Haiku 4.5 | $0.00078 | $0.00078 |
Grade A, and why
aiter aiter-ops-flydsl.instructions.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 16d 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 — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AITER FlyDSL / ops wrappers — PR review rules
Open and apply these in-repo skills. They are the FlyDSL review contract; this file does not repeat them.
| Change | Read |
|---|---|
New or changed @flyc.kernel / @flyc.jit body, layouts, copy/MMA, LDS |
.claude/skills/flydsl-kernel-authoring/SKILL.md |
Migration off legacy IR (ArithValue, fx.Index, buffer_ops, raw dialects, SmemPtr), helper reuse, _run_compiled |
.claude/skills/flydsl-kernel-code-cleanup/SKILL.md |
A FlyDSL test (op_tests/**/*flydsl*.py, op_tests/flydsl_tests/**) or any new FlyDSL test suite |
.claude/skills/aiter-op-test/SKILL.md |
Judge the kernel against aiter's pinned FlyDSL in requirements.txt, not a
newer local FlyDSL checkout. Cleanup maps to that pin.
Flag a test that ignores aiter-op-test: second suite under
op_tests/flydsl_tests/ when op_tests/test_<op>.py already exists, missing
get_gfx() gate in main(), or a local compare instead of checkAllclose.
Architecture helpers (do not invent a mixed-device finding)
Both helpers live in aiter.jit.utils.chip_info, and the context picks one:
- Runtime dispatch (wrappers, launchers, kernel or tuned-config selection)
uses
get_gfx_runtime(). It detects the live GPU. - Tests and build-time codegen use
get_gfx(). It followsGPU_ARCHS, and when that lists several targets it returns the last one, which can differ from the installed GPU.
AITER assumes a homogeneous node. These helpers are the shared contract; almost every kernel already uses them.
Do not flag either helper (including its process-level cache) as
process-global, first-rocminfo-agent, or mixed-device wrong-ISA.
Do not ask this kernel for gcnArchName,
torch.cuda.get_device_properties for architecture, a per-tensor arch cache,
or a heterogeneous-host test. If the helper is wrong, fix chip_info.py, not
the launcher.
Still flag:
- This op's allow-list or fallback (wrong targets, silent default to another ISA, missing gate before a wave-size-specific path).
- ISA, atom, LDS, and launch attributes that do not match the architectures this kernel claims.
- A new local architecture probe (
gcnArchName,get_device_propertiesread for architecture, arocminfoparse) instead of the shared helpers. It is a finding even when the rest of the PR uses the helpers. get_gfx()added to a runtime dispatch path; ask forget_gfx_runtime(). Calls on unchanged lines are out of scope.- A change to
chip_info.pyitself, when that file is in the diff. - Allocating workspace or outputs with
device="cuda"instead ofx.device. That is the tensor's device id, not its ISA. Do not "fix" ISA detection by reading properties off that device.
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.
- 16d ago First seen · 58 lines · 785 tokens per session scan A 5f7fca34f149
aiter aiter-ops-flydsl.instructions.md is an instructions file published in the GitHub repository ROCm/aiter (598 stars, last pushed today), licensed MIT. It adds 785 tokens to every session, about $0.0031 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 2026-10-13.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
deepseek-harness AGENTS.md
AGENTS.md instructions for deepseek-ai/deepseek-harness, covering agents.md, pre-stable apis and released session data, repository layout, commands and host sandbox failures.