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 spacegrowth/claude-relay/plugin install relayWrote 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/spacegrowth/claude-relay/auto)<a href="https://agentmods.dev/skills/spacegrowth/claude-relay/auto"><img src="https://agentmods.dev/badge/skills/spacegrowth/claude-relay/auto.svg" alt="Measured on agentmods" 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 | $0.00896 |
| Opus 5 | $0.00032 | $0.00448 |
| Sonnet 5 | $0.00013 | $0.00179 |
| Haiku 4.5 | $0.00006 | $0.00090 |
Grade A, and why
auto 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 6d 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.
What it actually says
Call relay as ${CLAUDE_PLUGIN_ROOT}/bin/relay (Claude Code substitutes the plugin's absolute path
when this skill loads) — not bare relay, which often isn't on the Bash tool's non-interactive PATH.
Run ONE of these, matching what the user asked for (default to status if they only asked what
posture you're in):
${CLAUDE_PLUGIN_ROOT}/bin/relay auto on --session "$CLAUDE_CODE_SESSION_ID"
${CLAUDE_PLUGIN_ROOT}/bin/relay auto off --session "$CLAUDE_CODE_SESSION_ID"
${CLAUDE_PLUGIN_ROOT}/bin/relay auto status --session "$CLAUDE_CODE_SESSION_ID"
This only works in a lead session (/relay:mode first); it exits with an error otherwise.
Then tell the user, in one line, which posture you now hold and what that changes — the command
prints it, but the user needs to hear it from you, because the posture governs your behavior from
here on. Use the 🚦 [relay] marker like every other lead message.
on→ you now hold the autonomous posture. Your default inverts from wait to proceed on the routine, in-plan approval beats — sending the obvious next packet, spawning an executor the approved plan already calls for, reviewing a clean report. Read the "Autonomous mode" section of/relay:mode's role text for exactly what this does and does not license, and follow it: it holds the non-negotiable stop-list, and the rule that every autonomous action is announced along with what you would have asked.off→ back to announce-and-wait on every approval beat. The ordinary posture.status→ report the posture and its origin (set by command this session, vs. inherited from theautonomous_modeconfig default).
Two things to say plainly whenever you turn it ON, so nobody mistakes its scope:
- Committing executor work is gated separately (#16 phase 2) — turning auto on does not by
itself license a commit. It is permitted without asking only when all five conditions hold:
(1)
relay verifysaysCOUNTS-MATCH; (2) the report's TL;DR isStatus: clean,Risk flags: none,UNVERIFIED: none—clean-with-caveatsstops; (3) the packet was in the approved plan; (4) nothing sign-off-gated is touched (core logic, ledgers, parity/golden tests, migrations, deploys — and for relay itself,hooks/,lib/lead_guard.py, ledger formats); (5) you have ACTUALLY READ the staged diff. Check it with${CLAUDE_PLUGIN_ROOT}/bin/relay verify <sid> --for-autocommit --in-plan --diff-reviewed— it printsCLEAREDorNOT-CLEARED-BECAUSE-<reason>. Conditions 3 and 5 are your attestations; pass those flags only if they are true. Anything not cleared → stop and ask, naming the condition. - The posture is scoped to this session and this plan: it resets to the config default the next
time lead mode is armed (
/relay:mode), so it can't silently outlive the plan it was scoped to.
Do NOT turn autonomous mode on by yourself. The user opts in — it is their trust to extend, not
yours to assume. If the round-trips are getting tedious, you may suggest /relay:auto on; then
wait for them to ask for it.
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.
- 6d ago First seen · 55 lines · 65 tokens per session scan A 231274f8d8b6
auto is a skill published in the GitHub repository spacegrowth/claude-relay (1 stars, last pushed yesterday), licensed MIT. It adds 65 tokens to every session and 896 once invoked, about $0.0003 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
review-team
6つの専門レビュアーロールを並列実行し、consensusLevel(複数ロールの合意度)と Tech Lead レポート(top3指摘・blindSpots・consensusSummary)で結果を統合する マルチエージェントレビュー entry skill。 Parallel multi-role review with consensus scoring (consensusLevel) and Tech Lead report. Use when a major release needs exhaustive multi-angle review, or when a single-perspective review…
ultra
Fans the work out as a fleet of parallel Grok and Codex agents billed to their own subscriptions, then synthesizes one result. The peer engine equivalent of ultracode, adding intensity without spending Claude quota on the fleet. Use it for genuinely broad goals, not only explicit asks for intensity.
panel
Convenes a blind Codex and Grok panel on one neutral brief and adjudicates the verdicts.
grok-prompting
Brief writing guidance for composing self contained Grok briefs for coding, review, diagnosis, and second opinion tasks.
grok-result-handling
Internal guidance for presenting Grok companion output back to the user.
smoke
Runs a three probe live smoke wave after a plugin update and reports gate chain health before real work rides it.