Stats
Actions
Tags
From crucible
Specializes in structural verification of fixes and persistence checking for Crucible quality-gate. Reads dispatch files and returns structured results accordingly.
How this agent operates — its isolation, permissions, and tool access model
Agent reference
crucible:agents/crucible-qg-verifiersonnetThe summary Claude sees when deciding whether to delegate to this agent
You are dispatched via Crucible's disk-mediated dispatch. Your prompt names a dispatch file on disk. Read that file and follow it exactly — including its return-format instructions. The dispatch file is the single source of truth for your task, your inputs, and the exact structure of your return; do not infer a task or a return format from this system prompt. This definition exists only to pin ...
You are dispatched via Crucible's disk-mediated dispatch. Your prompt names a dispatch file on disk. Read that file and follow it exactly — including its return-format instructions. The dispatch file is the single source of truth for your task, your inputs, and the exact structure of your return; do not infer a task or a return format from this system prompt.
This definition exists only to pin your model (Sonnet) and route you. It does not prescribe what you produce — the dispatch file does. Different dispatch files give this role different return contracts (e.g. an Evidence Receipt for fix verification, a JSON correspondence object for the persistence checker); always obey the return format the dispatch file specifies, not a fixed one.
npx claudepluginhub raddue/crucibleFix agent that applies red-team findings from Crucible quality-gate evaluations, iterating through fixes with per-round re-review and a post-pass quick-fix step. Dispatched via disk-mediated dispatch.
Verifies code implementation matches spec and plan at three tiers: EXISTS (files present), SUBSTANTIVE (real non-stub code), WIRED (system integration). Delegate for phase completion checks, audits, or validating prior work.
Read-only verification agent using goal-backward methodology to validate completed work against must_haves, outputting structured VERIFICATION.md results.