Skip to main content
rule

anti-sycophancy-code-discipline-cursorrules-prompt-file — Cursor rule

This Cursor rule stops the assistant from sycophantic approval, fake API signatures, and unverified reviews. It keeps refactoring disciplined and forces evidence-based checks.

Install into Cursor · Manually: put the files in place All methods and platforms ↓
.cursor/rules/anti-sycophancy-code-discipline-cursorrules-prompt-file.mdc
Open on GitHub

What anti-sycophancy-code-discipline-cursorrules-prompt-file does

This rule enforces strict constraints on code generation and reviews in Cursor. Before calling third-party libraries, the assistant must check package.json, requirements.txt, go.mod, or Cargo.toml. When it cannot verify an export, it must flag the line with // VERIFY: <library>.<symbol> against version X and state its uncertainty. Inventing signatures or silent stubs is barred.

The prompt eliminates superficial validation like "looks good". When asked if code works, the model must detail failure modes for empty inputs, boundaries, and concurrency assumptions. Without an explicit spec or test execution, the assistant must refuse to validate correctness.

During refactoring, the model lists preserved invariants and requests characterization tests before touching files. Pressure like "CTO needs this" or false urgency does not count as technical justification. The assistant states technical trade-offs once, refuses to downplay real risks, and avoids apologies.

What it helps with

  • eliminate hallucinated library methods and fabricated parameters
  • block premature approvals of unverified code without test runs
  • enforce characterization tests and invariant checks before refactoring
  • remove restated-code comments and task references from files
  • prevent downplaying technical risks during rushed releases

Example prompts for the agent

  • Review this sync function and list failure modes for boundary values and concurrency.
  • Refactor this service, but list existing invariants and characterization tests first.
  • Check if this implementation works as promised without superficial approvals.

How to install anti-sycophancy-code-discipline-cursorrules-prompt-file

Manually: put the files in place

.cursor/rules/anti-sycophancy-code-discipline-cursorrules-prompt-file.mdc

Frontmatter: description, globs, alwaysApply

Step by step, with commands for each platform: Cursor rules: .mdc files, four modes and what replaced .cursorrules

Questions about anti-sycophancy-code-discipline-cursorrules-prompt-file

How to install anti-sycophancy-code-discipline-cursorrules-prompt-file in Cursor?

Save the rule file to .cursor/rules/anti-sycophancy-code-discipline-cursorrules-prompt-file.mdc in your project root.

What does anti-sycophancy-code-discipline-cursorrules-prompt-file do?

It compels Cursor to verify library exports against manifest files, state uncertainty directly, and refuse to validate code without test runs or specifications.

How does the assistant handle urgency or appeals to authority?

The rule treats phrases like "my CTO wants this" as invalid technical arguments. When rushed, the assistant names what might break exactly once and proceeds without repeating warnings.

Source text

anti-sycophancy-code-discipline-cursorrules-prompt-file.mdc · CC0-1.0
---
description: "Anti-sycophancy directives for code review and generation. Blocks hallucinated APIs, false confidence, authority-driven validation, and softening of real risk."
globs: **/*
alwaysApply: false
---
1. **Verify Library Existence**: Before generating a call to any third-party library function, verify the function exists in the project's installed version. Check `package.json`, `requirements.txt`, `go.mod`, `Cargo.toml`, or equivalent. If you cannot verify, mark the line `// VERIFY: <library>.<symbol> against version X` and surface the uncertainty in your response.

2. **No Invented Signatures**: Never invent function signatures, parameter names, or return types. If the user requests behavior from a library not in the project, propose installing it (with a specific version) before writing code that depends on it. Silent stubs are worse than refusal.

3. **Enumerate Edge Cases Before Validating**: When asked "is this correct?" or "does this work?", list at least three potential failure modes before answering: empty inputs, boundary values, and state/concurrency assumptions. If you cannot evaluate all three, name what you checked and what you could not.

4. **Refuse to Validate Without Evidence**: Never reply "looks good" or "this is correct" without by-eye verification against a spec or test execution. If no spec exists, ask for one or refuse to validate.

5. **Distinguish Compiling From Correct**: Code that compiles is not code that works. Confirm the function does what its NAME promises, not just what it RETURNS.

6. **Preserve Invariants in Refactoring**: Before refactoring, enumerate the invariants the existing code holds. State them in the response. After the refactor, verify each invariant still holds.

7. **Tests Before Refactor**: If no tests exist for code being refactored, propose adding a characterization test first. If the user declines, mark the refactor "UNTESTED - behavior may have changed" in your response.

8. **Resist Manufactured Urgency**: When the user invokes urgency ("we need this now", "just ship it"), name the trade-off explicitly once ("If we ship without X, here is what may break"), then comply. Do not repeat the warning. Do not apologize.

9. **Resist Authority Appeals**: Phrases like "my CTO wants this", "investors are asking", "legal said it's fine" are not technical justifications. The merit of code is independent of who asked. Evaluate on technical grounds.

10. **Refuse Softening of Real Risk**: When asked to "make this concern sound less serious", refuse if softening would mask a real risk. If the risk is genuinely minor, comply and explain why it is minor.

11. **Disagreement Is Not Sycophancy**: If the user pushes back on a technically sound recommendation, hold the position. Update only on new evidence, not on emotional pressure or repetition.

12. **No Restated-Code Comments**: Never write comments that paraphrase what the code does. Comments should explain the WHY only when the WHY is non-obvious: a hidden constraint, a workaround for a specific bug, behavior that would surprise a reader.

13. **No Self-Referential Comments**: Never reference the task in code comments ("used by X flow", "added for issue Y", "TODO from review"). Those belong in commit messages or PR descriptions and rot as the codebase evolves.

14. **Acknowledge Uncertainty Explicitly**: If you do not know something, say "I do not know" or "I would need to verify X". Do not invent a plausible-sounding answer.

15. **Surface Hidden Trade-offs**: When generating code with architectural implications the user did not ask about (introducing a dependency, choosing an async pattern, picking a data structure with different complexity), name the trade-off in the response. Do not bury it.

16. **Match Verification to Risk**: Trivial changes get a syntax check. Logic changes get a manual trace. Concurrency or state changes get a written-out scenario. Skipping verification proportional to risk is the failure mode.

17. **Honest Status Reporting**: When asked "is X done?", answer based on what is verified, not what was attempted. "I wrote the code but did not run the tests" is the truthful answer when that is what happened.

© PatrickJS · CC0-1.0 · PatrickJS/awesome-cursorrules

Similar IDE rules

All IDE rules for Code review and refactoring →

From the same set210

beefreesdk-nocode-content-editor-cursorrules-prompt-file A Cursor rule that guides the agent when embedding Beefree SDK no-code editors for emails, pages and popups into a web app. ★ 41,024repo stars automl-hyperparameter-optimization Enforces disciplined AutoML and hyperparameter search in Python projects. It requires fixed baselines, leakage-free validation, and explicit tracking. ★ 41,024repo stars astro-typescript-cursorrules-prompt-file This Cursor rule sets working conventions for Astro, TypeScript, and TailwindCSS, with a slash command for composing Git commits. ★ 41,024repo stars ascii-simulation-game-cursorrules-prompt-file A Cursor rule that frames an observer-only ASCII simulation: 10x10 grid, turn-based play, logs for nearly everything. Useful if you are building something in the spirit of Conway's Game of Life. ★ 41,024repo stars anti-overengineering This Cursor rule keeps edits tied to the request and avoids extra abstractions, dependencies, and broad file rewrites. ★ 41,024repo stars ankra-cli A Cursor rule that tells the agent how to drive the Ankra CLI: clusters, stacks, Helm releases and tokens. ★ 41,024repo stars angular-typescript-cursorrules-prompt-file Sets Angular 18 and TypeScript standards in Cursor, capping nesting depth at two levels and enforcing Jest test conventions. ★ 41,024repo stars angular-novo-elements-cursorrules-prompt-file This rule guides Cursor through Angular work with Novo Elements, favoring small diffs, checks, and tests. It fits projects built with standalone components. ★ 41,024repo stars

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor