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.
npx agentmods add commands/gitguffaw/codex-router/execgit clone --depth 1 https://github.com/gitguffaw/codex-routerWhat 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 | $0.00011 | $0.00610 |
| Opus 5 | $0.00005 | $0.00305 |
| Sonnet 5 | $0.00002 | $0.00122 |
| Haiku 4.5 | $0.00001 | $0.00061 |
Grade A, and why
exec 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 yesterday.
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.
What it actually says
Run a Codex Router exec job through the shared runtime.
Raw slash-command arguments:
$ARGUMENTS
Core constraints:
- This is the only policy-routed analyze/exec write-capable entrypoint (
mode=exec, sandbox workspace-write). /codex-router:rescueis a separate task path: it may write only when the rescue subagent forwardstask --write(default for fix work; read-only when the user asks for diagnosis-only)./codex-router:cliis a raw Codex CLI escape hatch and is not job-tracked write routing.- Keep all user prompt text inside the companion runtime arguments; do not reinterpret or implement it in Claude.
- The companion runtime records the selected policy, mode, modifiers, and model controls in a context pack.
--search,--docs,--tool, and--parallelare Codex-side routing directives. Preserve them exactly; the companion runtime turns them into explicit inner-Codex instructions.-c/--config,--enable, and--disableare Codex config controls. Preserve them exactly.--waitand--backgroundare mutually exclusive. If both are present, stop with an error and do not invoke the companion runtime.
Foreground flow (the default, and whenever --wait is present):
node "${CLAUDE_PLUGIN_ROOT}/scripts/codex-companion.mjs" exec "$ARGUMENTS"
Return the command stdout verbatim, exactly as-is.
Background flow (only when --background is present):
- Run the same companion command in the foreground. It detaches the Codex worker and returns a launch stub containing the exact job id.
- Extract that exact job id from stdout.
- Launch the completion watcher as a Claude background Bash task:
Bash({
command: `node "${CLAUDE_PLUGIN_ROOT}/scripts/codex-companion.mjs" await-result "${jobId}"`,
description: "Codex execution completion",
run_in_background: true
})
- Do not call
BashOutputor wait for the watcher. Return the original launch stdout verbatim.
The watcher emits one concise terminal-status notification. It does not inject the full Codex result; /codex-router:result <job-id> remains the full-output surface.
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.
- yesterday First seen · 48 lines · 11 tokens per session scan A d94a0cac5787
exec is a command published in the GitHub repository gitguffaw/codex-router (2 stars, last pushed 4d ago), licensed Apache-2.0. It adds 11 tokens to every session and 610 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 commands, from other repositories
release
Prepare a new release for sandvault. This command updates the changelog, bumps the version, and creates a PR.
verify
Run the full verification pipeline before committing.
codex-review-waive
Waive a disputed Codex review finding so it stops blocking commits.
review
Review the current plan with peer LLMs via inspectrum.
codex-review
Run a Codex code review on commits, branches, or uncommitted changes.
cancel
Cancel an active background Codex job in this repository.