# Principle Least Privilege

> Limits tools, credentials, and environment to the task. Apply when granting capabilities to any role or child process.

- Skill: `bostonaholic/principle-least-privilege` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add bostonaholic/principle-least-privilege`
- Raw SKILL.md: https://api.skillmd.com/api/skills/bostonaholic/principle-least-privilege/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: bostonaholic (https://skillmd.com/u/bostonaholic)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/bostonaholic/principle-least-privilege

---


# Least Privilege

Enforce constraints by withholding the capability, not by asking for restraint; prompts never change a role's actual tools.

- Reviewers hold no Write/Edit and run in plan mode; they cannot fix and approve their own findings.
- Give child processes an environment allowlist and their own credential block, never another vendor's or the parent's full environment.
- Dispatch to the narrowest target able to perform the task.
- Refuse command-sink targets for required read-only guarantees whenever a narrower target is available.
- If every target exceeds the task's needs, scope the prompt explicitly and report the guarantee as prompt-level, not structural.
- When falling back to a full-tool context, state that the structural guarantee no longer applies.