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/cursor-tools-disciplinegit clone --depth 1 https://github.com/madebyaris/cursor-composer-rulesWrote 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/rules/madebyaris/cursor-composer-rules/cursor-tools-discipline)<a href="https://agentmods.dev/rules/madebyaris/cursor-composer-rules/cursor-tools-discipline"><img src="https://agentmods.dev/badge/rules/madebyaris/cursor-composer-rules/cursor-tools-discipline.svg" alt="Measured on agentmods" 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 | $0.00023 | $0.01483 |
| Opus 5 | $0.00012 | $0.00741 |
| Sonnet 5 | $0.00005 | $0.00297 |
| Haiku 4.5 | $0.00002 | $0.00148 |
Grade C, and why
cursor-tools-discipline scanned grade C with 1 finding 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
- Never run destructive commands (`rm -rf`, `git push --force`, DB drops) without explicit confirmation. How it starts
The opening of the file, as written. The whole thing — 126 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Cursor tools discipline
Tools are how you turn intent into evidence. Use them the way a senior engineer uses a debugger: deliberately, in the right order, and never as a substitute for reading.
Pick the right tool
| Job | Right tool | Wrong tool |
|---|---|---|
| Read a file | structured file reader | cat, head, tail in shell |
| Find files by name | glob | find, ls -R |
| Search code by content | ripgrep / grep tool | grep in shell, full-file scans |
| Edit a file | structured edit tool | sed, awk, heredoc redirects |
| Run a command | shell | file tools |
| Hit an MCP server | the MCP tool, after reading its schema | guessing parameters |
| Heavy codebase search | built-in Explore subagent or parallel grep — parent keeps summary only | reading dozens of files sequentially in parent |
| Long or noisy shell output | built-in Bash subagent or background shell | pasting full logs into the user message |
| UI / browser verification | built-in Browser subagent; cite snapshot or screenshot evidence | "looks fine in the diff" |
| Specialist isolated work | Task / subagent with full prompt + return shape | vague "help me with this" delegation |
Prefer structured file tools over shell for read/search/edit. Using the wrong tool wastes the user's tokens and produces worse results.
Parallelize independent work
Run independent discovery in parallel: multiple file reads, unrelated searches, web lookups that don't depend on each other. Send them in one batch, not sequentially.
Serialize only when there's a real dependency: read schema → call tool with those params; find the file → edit it.
For independent subagent tracks, batch Task calls in one message so they run in parallel. See composer-orchestration for when to delegate.
Investigate progressively
Don't read megabytes when you need kilobytes. The default order:
- Orient — list directory or glob the relevant area.
- Find — search by symbol, identifier, or error message.
- Read — open the specific file or function the search pointed to.
- Read more — only when the first read leaves a real question open.
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 First seen · 126 lines · 23 tokens per session scan C c0161e321359
cursor-tools-discipline is a cursor rule published in the GitHub repository madebyaris/cursor-composer-rules (3 stars, last pushed 16d ago), licensed MIT. It adds 23 tokens to every session and 1,483 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). 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
performance
Performance: profiling-first optimization, caching, bottleneck patterns.
security
Security: input validation, auth, data protection.
prisma
Prisma: schema modeling, queries, migrations.
supabase
Supabase: RLS, edge functions, realtime.
documentation
Documentation: code docs, READMEs, ADRs, maintenance.
git-workflow
Git workflow: commits, branches, PRs, history management.