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 rules/liuyihey/agent-engineering/request-decision-flowgit clone --depth 1 https://github.com/LiuYihey/Agent-EngineeringWhat 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 | $0.00387 | $0.00387 |
| Opus 5 | $0.00193 | $0.00193 |
| Sonnet 5 | $0.00077 | $0.00077 |
| Haiku 4.5 | $0.00039 | $0.00039 |
Grade A, and why
request-decision-flow 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.
What it actually says
Request Decision Flow
Default: follow this workflow for every user request unless it is clearly trivial or the solution is very deterministic. Do not try to judge whether a request is "complex" - when in doubt, apply this rule.
When to skip this rule
Skip only when both are true:
- Trivial problem - a single, obvious change with no meaningful trade-offs (typo fix, rename, one-line config tweak, pure explanation of existing code, "run this command").
- Very deterministic solution - exactly one correct approach; no real design choice (user already named the approach, or codebase convention leaves no alternative).
If either condition is unclear, do not skip.
Skill-first routing
Before implementation, pick the right skill:
- Requirements unclear, many constraints, or high-impact decisions -> call the
grill-meskill first and clarify with the user. - Design / exploration / discovery (architecture, new feature shape, trade-off analysis) -> call the
brainstormingskill first to widen and structure the solution space.
When unsure which skill fits, prefer grill-me for execution-bound tasks and brainstorming for open-ended design.
Multi-solution decision gate
If there are multiple valid approaches that could satisfy the requirement, call the AskQuestion tool and present clear options before implementation.
- Put the recommended option first and mark it as recommended.
- Do not bypass this step unless the user has already explicitly chosen an approach.
Scope control
- Keep options focused on the same user goal; do not expand into unrelated refactors.
- After user selection, execute decisively on the chosen path.
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 · 39 lines · 387 tokens per session scan A a08248b5a10c
request-decision-flow is a cursor rule published in the GitHub repository LiuYihey/Agent-Engineering (5 stars, last pushed 28d ago), licensed MIT. It adds 387 tokens to every session, about $0.0019 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 2026-08-31.
Other cursor rules, from other repositories
cuopt-skill-evolution
Trigger skill evolution — after solving non-trivial problems, propose skill updates.
cursorrules
See CLAUDE.md for full context. Quick rules.
moyu
当检测到过度工程模式时自动激活：修改用户未要求的代码、创建不必要的抽象层、 添加未要求的注释/文档/测试、引入不必要的依赖、重写整个文件而非最小编辑、 diff 超出请求范围。用户说"太多了"、"只改 X"、"简单点"时触发。 Activates on over-engineering: modifying unrequested code, unnecessary abstractions, unsolicited comments/docs/tests, unrequested dependencies, full file rewrites, scope-exceeding diffs. Triggers on "too…
cursorrules
This repository contains a structured skill library for AI-assisted product development. Skills are in skills//SKILL.md. Read the relevant SKILL.md before generating any artifact.
cursorrules
AgentAnycast is a decentralized P2P runtime for the A2A (Agent-to-Agent) protocol.
homunculus
Homunculus self-evolving AI assistant. Use when user asks about evolution cycles, goal trees, system status, instincts, or skills. Provides routing to detailed evolution workflows.