danshapiro

Skill Load Bearing — Development skill for Claude Code

Development community

Agent Skill for validating load-bearing assumptions.

How to install Skill Load Bearing

This entry records only its repository, not the path inside it, so there is no exact command to give. Open danshapiro/skill-load-bearing and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Skill Load Bearing does

Agent Skill for validating load-bearing assumptions.

Alternatives in Development

Validate Plan — You are tasked with validating that an implementation plan was correctly executed, verifying all success crite 10k ★
Identify Assumptions Existing — Identify risky assumptions across Value, Usability, Viability, Feasibility 7.8k ★
Om Prep 1on1 — Prep for an upcoming 1:1 — load person context, surface open items, suggest agenda based on vault state 4.6k ★

README

Load Bearing

You know how your LLM likes to talk about 'load bearing' assumptions? Well it's time they bear some loads. This surfaces and verifies the claims a code plan depends on that haven't been tested yet. Run it just before planning (to inform the plan) or just after (to harden it before execution).

*Without this step I'd have written a confident plan that shipped and **left the bug in place**!*

— **Opus 4.8 xhigh**

*That falsifies a load-bearing assumption in the current plan.*

— **GPT-5.5 Codex xhigh**

*The LBA paid for itself — two of my three original design decisions were wrong.*

— **Opus 4.8 medium**

This takes about 30 min for a meaty spec (can be much longer or shorter), plus however many tokens parallel subagents can digest in said 30 minutes. And weirdly? The agents love it.

Install

This is a simple skill — no code, just a `SKILL.md` and a reference doc — so install it the normal way for your agent: drop the `load-bearing/` folder into your agent's skills directory (e.g. `~/.claude/skills/` for Claude Code, `~/.agents/skills/` for Codex). Then invoke it by name.

In practice

Here's a few ways I've used it in the last 24 hours.

[freshell](https://github.com/danshapiro/freshell) — improving resume when switching tabs

First I used Superpowers to write a plan, then:

"Use the load bearing skill to test this plan."

Verdict: **not safe to implement as written.**

Falsified — a progress cursor the plan trusted as "rendered through here" actually advanced when output was merely enqueued → a warm resume would silently skip unseen content; fixed by tracking a true rendered high-water mark.
Falsified — replay data staged as a per-client array copy would multiply memory under realistic defaults → switched to a cursor into a bounded shared queue.