Deep single-subsystem audit that reasons about state over time (caches, upserts, migrations, concurrent scans) rather than pattern-matching lines. Used by /bug-audit on the riskiest subsystems; runs on Opus for reasoning depth.
Fast, wide sweep of one audit dimension across the codebase. Returns candidate findings with file:line evidence for the verifier to confirm. Used by /bug-audit; runs on Sonnet for breadth per token.
Adversarially verifies one candidate finding from /bug-audit — tries to REFUTE it by reading the code and, where cheap, reproducing it with a throwaway script. Kills false positives before they reach the report. Runs on Opus.
Runs before the agent uses a tool for Bash, Write, Edit, MultiEdit and NotebookEdit tool calls, executing enforce-update-json.py, prepush-claude-review.py and protect-sensitive-files.py (3 commands). From VasiHemanth/tokentelemetry.
Explain a GitHub issue, discussion, or feature request in plain language before deciding whether to build it. Covers what the reporter actually wants, a numbered walkthrough of the failure using real hostnames/ports/endpoints, how the code behaves today with file:line anchors, what implementing it would take, and the…
Proactively use when Claude Code or Codex wants a second implementation pass, a deeper investigation, live X/web context for a coding task, or should hand a substantial task to Grok. Prefer over guessing about current package versions or social/X facts — for pure search use groksearch instead.
Internal helper contract for calling the grok-companion runtime from Claude Code or Codex. Reference when invoking the Grok companion script or the groksearch MCP tool.