zachcuffie

Sop Stress Test — Testing skill for Claude Code

Testing community

Claude skill that stress-tests SOPs, runbooks and checklists across business ops, cybersecurity and AI operations.

How to install Sop Stress Test

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

What Sop Stress Test does

Claude skill that stress-tests SOPs, runbooks and checklists across business ops, cybersecurity and AI operations. Produces a readiness verdict, complete findings register, traceable redlines, owner decision log and a revised draft — without inventing org facts. Beta; independent testers welcome.

Alternatives in Testing

Paperjury — Pre-submission AI review stress-test for research papers 1k ★
Workflows — Catalog of multi-agent workflows for development, testing, and business tasks 127 ★
Al Agent.Test.Prompt — Generate comprehensive test codeunits for Business Central Agent SDK integrations 109 ★

README

SOP Stress Test

A Claude skill that stress-tests standard operating procedures (SOPs), runbooks and checklists across **business operations, cybersecurity and AI operations**.

It reads a procedure the way it will actually be used: by a new employee following it literally, by the owner who answers for it, and by an operator under pressure. It reports what would make execution fail, then proposes fixes that trace back to those findings.

**Status: beta (v0.1.1).** The method and validation tooling are built and tested. How well it finds real defects is **not yet measured**: the only model runs so far were done by the author, with the answer keys visible. Independent testers are wanted; see [Contributing](CONTRIBUTING.md).

What you get

For each procedure reviewed:

Output	What it contains
Readiness verdict	One of: Insufficient evidence to assess, Not ready, Ready after non-blocking edits, No blocking defects identified within reviewed scope. This is document readiness only, never operational assurance.
Findings register	Every supported finding, with no top-10 cap. Each has its evidence, a failure scenario, where the reviewer looked for existing coverage, and a proposed fix.
Coverage ledger	Which checks were applied to which sections, with a short observation for each "no finding".
Decision log	Choices only the owner can make: roles, thresholds, time limits, tools, risk acceptance. These are marked, never invented.
Redlines	Before/after text edits. Each cites the findings it fixes, with a reason.
Revised SOP	A clean draft with unresolved decisions visibly marked. It is labeled not approved for operational use.

What makes it different

Separates what is known from what is not. Each finding is a confirmed defect, a conditional concern, an unverified dependency, or an optional improvement. Severity is rated separately from certainty. An unavailable policy is not