How do I install the Preflight skill?

Run npx skillmds@latest add zedarvates/preflight in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Preflight safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Preflight?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Preflight free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Preflight?

zedarvates (@zedarvates) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Preflight by zedarvates — Agent Skill

Name: Preflight
Author: zedarvates

Preflight

Make the token-saving optimizations automatic instead of opt-in — a committed project policy plus a UserPromptSubmit hook that injects the prefer-local rules and suggested skills on every prompt, so a fresh prompt (after a component update, or from another dev/agent) never bypasses local routing. Use when the user wants the prefer-local behaviour enforced automatically, consistency across multiple devs/agents, or asks why their prompts aren't using local models by default.

zedarvates 651f4f2 5 files · 11.4 KB Updated Sep 17, 2026

File contents

preflight — enforce prefer-local, automatically

The gap this closes: the optimizations (local routing, prompt improvement, skill search) are opt-in, so a fresh prompt — especially after an update or from another dev/agent — bypasses them. Enforcement that doesn't depend on the model remembering:

1. Committed policy (`.botte/policy.md`)

The shared source of truth every dev and agent reads: default to local for cheap work, escalate only hard reasoning, improve big prompts, run /checkup after updates, budget. Commit it so multi-dev / multi-agent stays consistent.

from skills.preflight import policy
policy.write_default(project)          # idempotent
policy.ensure_agents_pointer(project)  # links AGENTS.md/CLAUDE.md to it

2. Preflight hook (every turn, ~0 cost)

A UserPromptSubmit hook that, before the expensive model sees the prompt, injects: the prefer-local rule, the top relevant skills (lexical find_skills, 0 tokens), and an improve_prompt nudge for big/ambiguous requests. Crash-proof (a hook must never block you) and fast (no LLM call).

echo '{"prompt":"classify these tickets","cwd":"."}' | python -m skills.preflight.hook

The deployer wires it automatically into .claude/settings.json:

{"hooks": {"UserPromptSubmit": [{"hooks": [
  {"type": "command", "command": "python -m skills.preflight.hook"}]}]}}

Install

python -m skills.bootstrap.cli <project> writes the policy, wires the hook and points AGENTS.md at the policy — all non-destructively. Works in any harness with a pre-prompt hook (Claude Code today); the hook script is standalone.

Related: [[bootstrap]] (installs this), [[checkup]] (the canonical audit), [[skill_finder]], [[prompt_improver]], [[auto_router]].

---name: preflightdescription: Make the token-saving optimizations automatic instead of opt-in — a committed project policy plus a UserPromptSubmit hook that injects the prefer-local rules and suggested skills on every prompt, so a fresh prompt (after a component update, or from another dev/agent) never bypasses local routing. Use when the user wants the prefer-local behaviour enforced automatically, consistency across multiple devs/agents, or asks why their prompts aren't using local models by default.---# preflight — enforce prefer-local, automaticallyThe gap this closes: the optimizations (local routing, prompt improvement, skillsearch) are **opt-in**, so a fresh prompt — especially after an update or fromanother dev/agent — bypasses them. Enforcement that doesn't depend on the modelremembering:## 1. Committed policy (`.botte/policy.md`)The shared source of truth every dev and agent reads: default to local for cheapwork, escalate only hard reasoning, improve big prompts, run `/checkup` afterupdates, budget. **Commit it** so multi-dev / multi-agent stays consistent.```pythonfrom skills.preflight import policypolicy.write_default(project) # idempotentpolicy.ensure_agents_pointer(project) # links AGENTS.md/CLAUDE.md to it```## 2. Preflight hook (every turn, ~0 cost)A `UserPromptSubmit` hook that, before the expensive model sees the prompt,injects: the prefer-local rule, the top relevant skills (lexical `find_skills`,0 tokens), and an `improve_prompt` nudge for big/ambiguous requests. Crash-proof(a hook must never block you) and fast (no LLM call).```bashecho '{"prompt":"classify these tickets","cwd":"."}' | python -m skills.preflight.hook```The deployer wires it automatically into `.claude/settings.json`:```json{"hooks": {"UserPromptSubmit": [{"hooks": [ {"type": "command", "command": "python -m skills.preflight.hook"}]}]}}```## Install`python -m skills.bootstrap.cli <project>` writes the policy, wires the hook andpoints AGENTS.md at the policy — all non-destructively. Works in any harness witha pre-prompt hook (Claude Code today); the hook script is standalone.Related: [[bootstrap]] (installs this), [[checkup]] (the canonical audit),[[skill_finder]], [[prompt_improver]], [[auto_router]].

Frequently asked questions