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/verify)<a href="https://agentmods.dev/skills/spacegrowth/claude-relay/verify"><img src="https://agentmods.dev/badge/skills/spacegrowth/claude-relay/verify.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 | $0.00083 | $0.01688 |
| Opus 5 | $0.00042 | $0.00844 |
| Sonnet 5 | $0.00017 | $0.00338 |
| Haiku 4.5 | $0.00008 | $0.00169 |
Grade A, and why
verify 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.
How it starts
The opening of the file, as written. The whole thing — 110 lines — stays where its author put it; the contents beside it link to each section on GitHub.
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: ${CLAUDE_PLUGIN_ROOT}/bin/relay verify $session_id
Add --packet N for an older packet, and --rerun to also re-run the pytest-shaped commands the
report declares (off by default so verify stays fast).
READ THIS BEFORE YOU READ A VERDICT
A COUNTS-MATCH must never be read as "the report is true." This is the whole design
constraint, not a footnote. The field data behind it: across ~15 real executor reports, a
counts-verifier would have caught exactly one thing (a lint miss). Every dangerous problem
was premise-level — wrong oracle, wrong write-side, suite green in the wrong venv — and every one
of those is invisible to re-running the declared commands.
So: this tool checks that the numbers line up. It cannot check that the work was right. Your judgement on the staged diff stays the real check — if you let a COUNTS-MATCH displace it, you have made relay less safe, not more, because you have swapped the thing that caught the real problems for the thing that catches lint misses.
That is also why there is no "PASS", no "VERIFIED", and no "clean" in this command's vocabulary. The verdict words are deliberately chosen so they cannot be over-read at a glance.
The verdicts
| Verdict | Exit | Means |
|---|---|---|
COUNTS-MATCH |
0 | The checkable numbers agree. That is all it means. Still read the diff. |
MISMATCH |
1 | A claim contradicts staged reality — a file claimed but never staged, a "staged" confirmation over an empty index, or (with --rerun) a declared test count that doesn't reproduce. |
MALFORMED |
2 | The report violates the REPORT FORMAT's TL;DR contract — most often a missing UNVERIFIED: line. Per #6, an absent UNVERIFIED line reads as malformed, never as "nothing to report." |
INCONCLUSIVE |
3 | A check you asked for did not complete (e.g. --rerun produced no N passed count). Not a COUNTS-MATCH with a caveat — nothing was compared. |
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 · 110 lines · 83 tokens per session scan A a39395d5d6c5
verify is a skill published in the GitHub repository spacegrowth/claude-relay (1 stars, last pushed today), licensed MIT. It adds 83 tokens to every session and 1,688 once invoked, about $0.0004 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
A multi-reviewer code review process that checks a change from several specialist viewpoints and combines the results into one report. It can cover bugs, security, tests, dependencies, frontend behavior, and continuous-integration workflows.
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.