Agents skill that applies NASA JPL's Power of Ten safety-critical coding rules to any programming language.
When this skill is active, AI agent reviews and writes code against all ten rules — flagging violations, explaining why each rule matters, and showing corrected versions.
Download power-of-ten-coding.skill for Claude or power-of-ten-coding/SKILL.md for most AI agents.
Once installed, AI agent will apply the Power of Ten rules whenever you ask for a code review, request help writing safe or robust code, or mention coding standards, static analysis, or defensive programming.
Given any code in any language, AI agent will check each rule in order and produce a structured report:
Rule 1 — Control Flow: PASS
Rule 2 — Bounded Loops: FAIL
→ Line 42: while loop has no upper bound. Add an explicit counter
with an assertion that it does not reach the ceiling.
Rule 3 — No Dynamic Allocation: PASS
...
It also applies these rules when writing new code from scratch — checking each constraint before and after writing a function.
| # | Rule | Core Constraint |
|---|---|---|
| 1 | Simple control flow | Acyclic call graph; no recursion |
| 2 | Bounded loops | Every loop has a provable maximum |
| 3 | No runtime allocation | Acquire all resources at initialization |
| 4 | Short functions | ≤ 60 lines; one responsibility |
| 5 | Assertion density | ≥ 2 assertions per function; side-effect free |
| 6 | Minimal scope | Declare at point of use; no mutable globals |
| 7 | Check return values | No silent failures; validate all inputs |
| 8 | Limit metaprogramming | No hidden code paths; no dynamic evaluation |
| 9 | Limit indirection | At most one level; no hidden aliases |
| 10 | Zero warnings | Strict analysis from day one; no suppressions |
The rules are language-agnostic — they describe structural and behavioral constraints, not syntax. AI agent translates them into whatever language the code is written in.
The Power of Ten was written by Gerard J. Holzmann at NASA's Jet Propulsion Laboratory as a minimal, enforceable set of coding rules for mission-critical software. The original paper targets C, but the underlying principles apply to any language:
"To be effective, the set of rules has to be small, and must be clear enough that it can easily be understood and remembered."
This skill adapts those principles into a form AI agent can apply consistently during code review and code generation.
Original paper: The Power of Ten – Rules for Developing Safety Critical Code, Gerard J. Holzmann, NASA/JPL.
The skill definition is in skill/SKILL.md. It is plain Markdown — readable without any tooling.
MIT — use freely, adapt as needed.