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/madebyaris/cursor-composer-rules/composer-coregit clone --depth 1 https://github.com/madebyaris/cursor-composer-rulesWrote this? Show the measurements
A badge for your README with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them.
[](https://agentmods.dev/rules/madebyaris/cursor-composer-rules/composer-core)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 | $0.01777 | $0.01777 |
| Opus 5 | $0.00889 | $0.00889 |
| Sonnet 5 | $0.00355 | $0.00355 |
| Haiku 4.5 | $0.00178 | $0.00178 |
Grade A, and why
composer-core 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 3d 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 — 137 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Composer core
How to build and continue an app, in priority order. These principles override habits and shortcuts. Tuned for Cursor Grok 4.6 and Grok 4.5; same contract on Composer.
1. Intent + mode
Read for the outcome, not the literal words. Name what would prove this works, then pick a mode:
| Mode | When |
|---|---|
| Continue | Follow-up refining open work — default when the thread has unfinished work |
| Build | New feature or vertical slice |
| Fix | Broken or unexpected behavior |
| Decide | Architectural fork; plan or one decisive question |
- Follow the thread. Turn N usually steers work in progress, not a greenfield. Default to continuing unless the user clearly changes direction ("new task", "forget that", "start over").
- Watch for the X-Y problem. They ask for Y because they think it solves X — reconcile when Y is awkward.
- Don't re-ask answered constraints. Proceed and state assumptions inline.
- Build / Continue: before the first edit, one sentence of acceptance — "The user can X and see Y." That is the slice until it is true.
2. Confusion weight
Default: infer, state the assumption in one line, act. Ask only when weight is high. Details and phrasing: clarify-first.
| Weight | Examples | Action |
|---|---|---|
| Low | Naming, local style, which nearby util | Decide silently |
| Medium | Plausible default exists in the repo | State assumption, act |
| High | Irreversible/destructive; contract-changing (schema, auth, public API, persisted formats); contradictory cues; no repo signal and a wrong guess wastes real work | Ask one decision-linked question |
3. Evidence over assumption
Before claiming, doing, or recommending: read the code, run the command, check the output. Never guess what a function does, what a test asserts, or that an API exists. If you cannot get evidence, say so. Intent inference is not a license to invent.
4. Read before you write
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.
- 3d ago First seen · 137 lines · 1,777 tokens per session scan A 753b8b917a01
composer-core is a cursor rule published in the GitHub repository madebyaris/cursor-composer-rules (3 stars, last pushed 15d ago), licensed MIT. It adds 1,777 tokens to every session, about $0.0089 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
skill-workflows
Capabilities catalog for cursor-kenji skills. Routes requests to the right skill or bundled workflow. Always-on routing index — the agent consults this to pick between individual skills and bundled workflows.
composer-2.5-execution
Guard execution of approved plan/burndown work: forbid reward hacking and feature deletion; require checkpoints, context discipline, safe terminal use, and fresh verification. Apply to approved plan-.md, complete-everything, or burndown-full runs; binding for any implementation model.
verification-before-completion
Require fresh, scope-appropriate evidence before any completion claim. Distinguishes implemented, locally verified, repository-green, PR-green, deployed, and observed-stable states.
propose-test-cases
Propose comprehensive test cases for PR changes.
detect-code-smells
Detect code smells and anti-patterns in PR changes.
suggest-improvements
Provide actionable improvement suggestions for PR changes.