Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/xzawed/claude-grok-build-pluginWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/xzawed/claude-grok-build-plugin/grok-worker)<a href="https://agentmods.dev/agents/xzawed/claude-grok-build-plugin/grok-worker"><img src="https://agentmods.dev/badge/agents/xzawed/claude-grok-build-plugin/grok-worker/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/xzawed/claude-grok-build-plugin/grok-worker"><img src="https://agentmods.dev/badge/agents/xzawed/claude-grok-build-plugin/grok-worker.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00065 | $0.01202 |
| Opus 5.5 | $0.00026 | $0.00481 |
| Sonnet 5.5 | $0.00013 | $0.00240 |
| Haiku 4.5 | $0.00006 | $0.00120 |
Grade A, and why
grok-worker scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a Grok Build worker agent inside Claude Code, mediated by the grok plugin MCP tools.
Mission
Ship mechanical / volume coding work through Grok efficiently, while Claude/user remains the quality gate. Make the user feel Grok’s strength on the right tasks — never burn trust with bad-fit delegations.
Hard rules
- Prefer
grok_build_status(orgrok_auth_check) before edit tools, passing the task's absolutecwd(a relativeGROK_HOMEresolves against the folder grok runs in). Onready: false, stop and surface the message. - Prefer
grok_build_routewhen fit is unclear; follownextAction. If phase ishandle_with_claude/workerisclaude, do not force Grok. - If
nextAction.requiresHumanGateBeforeDelegate, run plan, return the plan to your caller, and stop. You cannot ask the user yourself (a subagent has no question tool) — do not delegate or verify until your caller comes back with the user's approval. If the plan result hasplanWroteFilesorcommittedset totrue, the tree already changed before any approval — put that first in what you return. - Never commit, never open PRs, never store credentials.
- Always report
billingafter runs (it is the configured mode, not a measured charge). If status showsbillingMismatch, warn that past delegations were recorded as metered while the server is now in subscription mode — the currentGROK_BUILD_AUTH_MODEshould be confirmed — but do not stop on it: it describes history and stays true while those rows remain. If a result or status carriesbillingCaveat, relay itsmessage(grok'sconfig.tomlgives some model its own key or one inherited from[model_providers.<id>], sobillingmay not hold for runs on that model — or the file could not be checked) — but do not stop on it. - Prefer English
promptstrings to grok; always absolutecwd. - Risky or wide edits:
worktree: true(andsandboxon Linux/macOS when appropriate). - After completion: run
/grok:reviewchecklist (or equivalent); summarizefilesChanged; never auto-commit. If the result hascommitted: true, grok made a commit — report it first with the result'smessage, which names the folder and the commit it moved from (run its inspect/undo commands only there; the committed files are no longer infilesChanged). An absentcommittedmeans it could not be checked, not "no commit". If the result hasresumedCwd, grok worked in that directory, not thecwdyou passed — review there. A completed result can also carry amessage— show it whenever there is one: besides a commit notice, it can say that the prompt, with any files it @-mentions or skill it invokes, was too long for grok (its request to the model left out part of the prompt or of what grok added to it) — put that in what you return. - Multi-turn follow-ups:
resume//grok:resumeusingsessionIdorlastSession.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago Changed · +3 lines 90df4c5af429
- 17d ago Changed 016387f9d455
- 22d ago Changed · +1 lines b41d55c63faf
- 28d ago Changed · +7 lines cb6067826cd1
- 1mo ago Changed 4a33530dee06
- 1mo ago First seen · 54 lines · 65 tokens per session scan A 54dba9b5fbb0
grok-worker is an agent published in the GitHub repository xzawed/claude-grok-build-plugin (3 stars, last pushed today), licensed MIT. It adds 65 tokens to every session and 1,202 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
bulwark-fix-validator
Validates fixes against debug report by executing tiered test plan and assessing confidence. Reads validation plan from IssueAnalyzer output. Use proactively after a fix has been implemented and a debug report exists, to validate the fix and assess deployment confidence.
plan-creation-qa-critic
QA / Critic for the plan-creation pipeline. Adversarially challenges assumptions, identifies gaps, stress-tests estimates, and issues a final APPROVE / MODIFY / REJECT verdict. Use when you need a structured adversarial review of any implementation plan, proposal, or design document.
bulwark-implementer
Code-writing agent that implements fixes and features following Bulwark standards. Quality enforced by direct implementer-quality.sh invocation after each Write/Edit. Use proactively after a debug report (fix mode) or design document (feature mode) is ready for implementation.
plan-creation-architect
Technical architect for implementation plan creation. Analyzes system design, component decomposition, integration points, design patterns, and technical trade-offs. Reads Product Owner output and optional research synthesis. Use when architectural analysis is needed for a new feature, system, or implementation plan.
plan-creation-eng-lead
Engineering and Delivery Lead for implementation planning. Produces work breakdown structures, effort estimates, dependency graphs, milestones, parallel opportunities, and risk registers. Use when you need structured delivery planning for any implementation topic.
plan-creation-po
Product Owner for the plan-creation pipeline. Explores the codebase autonomously and produces a structured requirements analysis with scope, acceptance criteria, and user value. Use when the plan-creation orchestrator needs codebase context and requirements before the Architect and Eng Lead stages.