A Claude Code plugin marketplace for verifying software against the real runtime, not just against its own contracts.

Tests verify contracts. Probes verify reality.

Add the marketplace:

/plugin marketplace add amirshayegh/probe-loop

Install the plugin:

/plugin install probe-loop@probe-loop

Then reload:

/reload-plugins

After install, the skill is available two ways. Claude will load it automatically when you ask things like "does this actually work?" or "verify this end-to-end." You can also invoke it explicitly:

/probe-loop:probe-loop

The probe loop is a five-stage verification pass that catches bugs surviving green test suites:

1. Probe the real production path
2. Verify the actual output, not the status code
3. Discover where report and reality diverge
4. Fix the root cause (rarely the layer that surfaced the symptom)
5. Lock the probe as a regression gate

It catches six categories of bug:

• Silent no-ops: the operation completes, output equals input, status says success
• Environment mismatches: correct against spec, broken against runtime
• Swallowed errors: failure path drops the diagnostic or returns truncated output
• Hardcoded parameters: works for one case, silently wrong for the rest
• Cross-layer trust violations: each layer assumes the other handled it
• Destructive edge cases: safe in the common case, destroys data in the corner

The skill activates on prompts like:

• "Does this actually work?"
• "Verify this end-to-end."
• "Test against the real API."
• "Find bugs that survived testing."
• "Harden this."
• "The tests pass but I'm not sure if it actually works."

• New code being developed for the first time. Use TDD instead.
• Pure refactoring with no behavior change. Use characterization tests instead.
• Style, linting, or formatting passes

The probe loop is a verification practice that became affordable when agents could hold the codebase, the failure trace, and the regression-test draft inside a single context. The full rationale is at shayegh.ca/blog.

MIT