2-4 clarification turns before the agent has enough scope to start useful implementation.
fela@agents:~$ @im "Fix login bug"
Agent preflight for AI coding agents
One-Click Improve
Before any coding agent starts, make the task routed, scoped, project-aware,
safe, and verifiable. @im and /im run preflight for Codex,
Claude Code, Cursor, VS Code, and terminal workflows.
@im terminal, Codex, VS Code
/im Cursor and Claude Code
automatic surface routing
root@one-click-improve: ~/install
curl -fsSL https://improv.sh/install.sh | bash
No key required
Best with your key
Local + model-backed
same prompt, different outcome
Stop paying agents to rediscover the task.
The value is not prettier wording. It is whether the first agent turn receives route, scope, project context, safety constraints, acceptance criteria, and validation evidence.
The task is preflighted up front with route-aware context, acceptance criteria, and validation.
For this rough `Fix login bug` request, before counting avoided rework and wrong-file edits.
without one-click improve
user: Fix login bug agent first response: - Which login flow is broken? - What error do you see? - What files should I inspect? - Should I add tests? cost: The session starts with discovery. The agent has no target flow, no done condition, no validation plan, and no surface-specific instructions. token burn estimate: ~1.5k-6k tokens before useful work.
with one-click improve
user: @im Fix login bug one-click output: ## Objective Fix the login bug in the affected auth flow. ## Context Confirm the failing login path, inspect the relevant auth files and nearby tests before editing, and preserve the current editor or workspace context. ## Constraints - Preserve auth API and session behavior unless the bug requires a focused change. - Avoid unrelated account or UI refactors. ## Deliverables - Code fix. - Tests for the failing login path. - Validation notes for valid and invalid credentials. ## Acceptance Criteria - Login succeeds for the affected flow. - Existing logout/session behavior does not regress. - The edge case is covered by a test. ## Validation Run targeted auth tests and record the manual login check. token burn estimate: ~200-700 prep tokens before useful work.
reproducible benchmarks
Prompt quality should be measured, not claimed.
The same fixtures used in CI publish a local benchmark report for route accuracy, required sections, safety signals, validation signals, and readiness score.
Generates benchmarks/prompt-preflight.json and a markdown report from deterministic fixtures.
Every target surface must include the right route, sections, risk language, and verification path.
Generated tasks carry lifecycle ids so the final agent result can be checked against the original criteria.
agent-prep loop
What happens before your agent sees the prompt
One-Click Improve runs preflight on intent: automatic route, repo context, constraints, deliverables, acceptance criteria, lifecycle id, and validation path.
$ invoke
Start from the panel, a selection, VS Code @im, Cursor /im, Claude Code /im, or terminal @im.
$ inspect
Detect missing scope, files, constraints, tests, ambiguity, repo signals, architecture notes, and done conditions.
$ preflight
Load the bundled skill, infer project context, and rewrite the request into an executable engineering task.
$ verify
Send or copy the result, then paste the agent final report back to check outcome evidence against the task.
agent operating layer
An agent preflight layer, not a thesaurus.
One-Click Improve sits between your intent and the coding agent. It normalizes a rough ask into the exact shape each agent surface can execute: command syntax, missing context, constraints, deliverables, acceptance criteria, validation steps, lifecycle tracking, and guardrails. The goal is higher yield from the same model call, with fewer correction loops.
improve-prompt skill as the source of truth for structure and task discipline.
works where prompts start
Two commands, every environment
The same preflight engine is exposed through @im and /im; routing happens underneath.
VS Code
Use the sidebar panel, improve selected text, or mention @im before a chat request.
@im Fix login bug
Cursor
Use the reliable slash command route for Cursor Chat and Cursor Agent workflows.
/im Add tests for session timeout
Terminal
Generate an improved prompt from any shell, then paste or pipe it into Claude Code or Codex.
@im "Refactor dashboard loading"
Explain Mode
Inspect the inferred route only when you want to see the machinery.
@im --explain "auth tests in Codex"
before and after
The output is designed for execution and verification
The improved prompt is not an essay. It is a work order with objective, context, constraints, deliverables, acceptance criteria, validation steps, and a lifecycle path.
Rough prompt
Fix login bug
Improved prompt
no key required
Works offline. Best with your key.
The local provider ships ready to use and now detects request type, domain, harness, repo context, safety, accessibility, and validation signals before it writes the task. Add your OpenAI, Anthropic, or OpenAI-compatible key when you want the strongest model-backed rewrite.
localNo network, smarter offline output
openaiSkill-guided model rewrite
anthropicClaude API rewrite
compatibleOpenRouter, Groq, Mistral, DeepSeek, Gemini, xAI
start here
Install once, preflight tasks everywhere
The installer detects VS Code and Cursor from your PATH or standard macOS app bundles, installs the extension, creates terminal shortcuts, writes Cursor and Claude command files, syncs the bundled skill to user-level harness folders, then optionally asks for a terminal provider key. You can skip it.
curl -fsSL https://improv.sh/install.sh | bash