# Principle Bounded Loops

> Requires explicit retry and watch limits. Apply when writing any loop, retry, polling, or wait cycle.

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

---


# Bounded Loops

Declare the bound with the loop: watch cycles, retries, poll budgets, helpers in flight, or a terminal verdict.

- Halt loudly at a count cap and report all unresolved state; never restart, extend, or soften exit criteria.
- Keep retry budgets small and stated: "retry once, then halt loudly".
- Treat a verdict as the bound; declare it and never supply a count the loop deliberately omits. The operator is the outer bound.
- Preserve Team's uncapped review verdicts: DESIGN ends on the reviewer's verdict; IMPLEMENT ends when no Blocking or Major finding remains.
- Give outputs explicit budgets: ~200-line design, ≤ 30-line helper reply, byte-capped prompt.
- Restructure over-budget output by dropping whole named units. Never silent truncation.