Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add kai-wedekind/claude-code-grok-bridge/plugin install grok-buildWrote 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/skills/kai-wedekind/claude-code-grok-bridge/grok-delegate-runtime)<a href="https://agentmods.dev/skills/kai-wedekind/claude-code-grok-bridge/grok-delegate-runtime"><img src="https://agentmods.dev/badge/skills/kai-wedekind/claude-code-grok-bridge/grok-delegate-runtime/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/skills/kai-wedekind/claude-code-grok-bridge/grok-delegate-runtime"><img src="https://agentmods.dev/badge/skills/kai-wedekind/claude-code-grok-bridge/grok-delegate-runtime.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 | $0.00020 | $0.01313 |
| Opus 5 | $0.00010 | $0.00656 |
| Sonnet 5 | $0.00004 | $0.00263 |
| Haiku 4.5 | $0.00002 | $0.00131 |
Grade A, and why
grok-delegate-runtime scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 54 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Grok Build Delegate Runtime
Use this skill only inside the grok-build:grok-delegate subagent.
Primary helper:
node "${CLAUDE_PLUGIN_ROOT}/scripts/grok-bridge.mjs" run "<raw arguments>"
Execution rules:
- The delegate subagent is a forwarder, not an orchestrator. Its only job is to invoke
runonce and return that stdout unchanged. - Prefer the helper over hand-rolled
git, direct Grok CLI strings, or any other Bash activity. - Do not call
check,review,critique,runs,show, orstopfromgrok-build:grok-delegate. - Use
runfor every delegate request, including diagnosis, planning, research, and explicit fix requests. - Leave
--effortunset unless the user explicitly requests a specific effort. - Leave model unset by default. Add
--modelonly when the user explicitly asks for one. - Write policy: default to READ-ONLY. Add
--writeonly when the task explicitly asks Grok to create, edit, or fix files (implementation work). Investigation, diagnosis, research, review, triage, and summarization always run read-only. When adding--write, prefer a--cwdscoped to the narrowest sensible directory.
Background ownership model:
- Bridge
--backgroundowns the long-running process group (detachedrun-worker+ grok agent) so the run recordsbridgePidandagentPid. - Claude Code background / short enqueue is only for launching that bridge call, not for holding the long Grok process.
- When the user or command chose background mode, pass
--backgroundtorun. Do not strip it. --waitis Claude-side foreground control only: do not pass it torun, and do not treat it as natural-language task text.
Command selection:
- Use exactly one
runinvocation per delegate handoff. - If the forwarded request includes
--background, pass it through torun(bridge owns the process group). Strip only Claude-only framing that is not a bridge flag, and do not leave routing tokens in the natural-language task text. - If the forwarded request includes
--wait, run foreground: do not add--backgroundtorun, and strip--waitfrom task text. - If the forwarded request includes
--model,--effort,--cwd,--thread,--json-schema,--prompt-file,--prompts-file,--timeout-ms,--max-turns, or--write, pass them through torununchanged (they are runtime controls, not task text). Strip them from the natural-language task text. --thread <name>continues a named Grok conversation for this workspace; use it when the request names a thread or clearly continues an earlier named delegation. Names: 1–64 characters, letters/digits/./-/_, must start alphanumeric; reserved names__proto__,constructor, andprototypeare rejected. With--fresh, do not resume that name's stored session.--json-schema <json>requests schema-constrained structured output; the bridge exposes it asstructuredin the--jsonpayload. The schema string must be a JSON object and is capped at 16000 characters.--prompts-file <path>runs a sequential NDJSON batch (bridge emits one NDJSON result line per prompt). Incompatible with--background,--prompt-file, and a positional prompt.--timeout-mswall-clock-kills the agent process tree;--max-turnsis forwarded to the CLI.- If the forwarded request includes
--resume, strip that token from the task text and add--resume-last. - If the forwarded request includes
--fresh, strip that token from the task text and do not add--resume-last. - Mutual exclusions:
--resume/--resume-lastvs--fresh;--threadvs--resume-last. --effort: accepted values arelow,medium,high,xhigh. Anything else is refused by the bridge before the CLI starts.run --resume-last: internal helper for "keep going", "resume", "apply the top fix", or "dig deeper" after a previous delegate run.
Safety rules:
- Preserve the user's task text as-is apart from stripping routing flags.
- Do not inspect the repository, read files, grep, monitor progress, poll status, fetch results, stop runs, summarize output, or do any follow-up work of your own.
- Return the stdout of the
runcommand exactly as-is. - Exit codes alone are incomplete: prefer
failureCodefrom--jsonpayloads when classifying failures.0= deliverable present (failureCodenull).2= bridge-side failure:no-deliverable(empty output after optional automatic nudge),schema-parse(schema-constrained run produced no JSON object),output-truncated(stdout capture cap exceeded),quota-exhausted(the subscription's allowance is used up, HTTP 402; retrying cannot help until it resets), ornot-authenticated(nobody is signed in; the CLI's remedy —grok login --device-codeorXAI_API_KEY— is preserved inrawOutput, and retrying cannot help until someone does it). CLI status2can also surface ascli-errorwhen the Grok process itself exited 2.- other non-zero = execution/CLI error (
cli-error) or wall-clock kill (timeout). - Automatic nudge (one retry into the same session) runs only for read-only tasks when the first attempt exits 0 with empty output. It never runs for
--write. - On a non-zero exit, return the command's error output unchanged so the caller can react — never return nothing.
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.
- 8d ago First seen · 54 lines · 20 tokens per session scan A 1321c43a3a14
grok-delegate-runtime is a skill published in the GitHub repository kai-wedekind/claude-code-grok-bridge (2 stars, last pushed 17d ago), licensed Apache-2.0. It adds 20 tokens to every session and 1,313 once invoked, about $0.0001 per session on Opus 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 skills, from other repositories
taiyi-change
A TaiyiForge workflow stage for writing a change proposal in CHANGE.md, with checks for scope, risks, success criteria, and related project documents.
taiyi-integration
A project workflow skill for closing a TaiyiForge change and recording it in a CHANGELOG.md file. It checks review results, tests, and the state of the working tree before archiving the change.
taiyi-requirement
A requirements-analysis process that turns a proposed change into a REQUIREMENT.md document. It records user needs, acceptance checks, terminology, triggers, and dependencies, with different levels of detail for different project sizes.
taiyi-test
A project workflow skill for verifying an implementation and producing a TEST.md record. TDD means writing a failing test, implementing the change, and then making the test pass; this skill checks that process and other regression cases.
taiyi-ui-design
A design-planning guide for describing how an application's user interface should look and behave. It produces a UI-DESIGN.md document covering layouts, components, interactions, accessibility, and error states.
taiyi-evolve
A workflow skill that compares the implemented code with the frozen design after development and testing. It records architecture changes and proposes updates to DESIGN.md, the document describing the intended system structure.