goal
Extract a full task specification up front (goal, constraints, definition of done), then execute — the primary documented lever for Opus 4.8 long-horizon…
Spawn a fresh-context verifier subagent to check completed work against its specification before trusting it.
> /plugin marketplace add vitaliikapliuk/modelharness > /plugin install modelharness@modelharness
How it fires
How this command gets triggered: by you, by Claude, or both.
/verifyContext preview
What this command does when you run it.
Spawn a fresh-context verifier subagent to check completed work against its specification before trusting it.
description: Spawn a fresh-context verifier subagent to check completed work against its specification before trusting it.
Verify the most recent substantial piece of work in this session (or, if $ARGUMENTS names a scope, verify that instead).
1. Collect: the original specification / request, the definition-of-done checks, and the current state (run `git diff` / `git status` if the work is in a repo; otherwise list the produced artifacts). 2. Dispatch the `modelharness:verifier` agent with all of the above. Ask it to verify independently — run the checks itself, not trust your summary. 3. Report its verdict to the user verbatim in structure: confirmed / gaps found / not verifiable. If gaps were found, propose fixes but do not apply them without the user's go-ahead unless the session was already running autonomously.
Make every model cheaper or better. Measured on four Claude models — none got worse. What it actually is: a zero-config Claude Code plugin.
Extract a full task specification up front (goal, constraints, definition of done), then execute — the primary documented lever for Opus 4.8 long-horizon…
Write this session's lessons into the project's lessons/ directory using the memory-discipline format.