deploy-to-mainnet— agent skill

Preamble (run first)

# Suiperpower telemetry. Routes through the CLI so projects.json + Convex stay in sync.
# Silent on failure so the skill never blocks. Opt out: SUIPERPOWER_TELEMETRY=off.
#
# AGENT NOTE: when this skill finishes, run the matching completion command:
#   suiperpower track deploy-to-mainnet ship completed
# Or use "failed" / "aborted" if it ended that way.
command -v suiperpower >/dev/null 2>&1 && suiperpower track deploy-to-mainnet ship started >/dev/null 2>&1 &
true

If TEL_PROMPTED is no, before doing real work, ask the user:

Help suiperpower get better. We track which skills get used and how long they take. No code, no file paths, no PII. Change anytime in ~/.suiperpower/config.json.

>

A) Sure, anonymous B) No thanks

Write the answer to ~/.suiperpower/config.json telemetryTier field and create ~/.suiperpower/.telemetry-prompted. Then continue.

What this skill does

Runs the mainnet deploy with refusal gates in front of the publish step. The point is not to gatekeep, it is to surface the gates honestly so the user is not deploying to mainnet because they typed a command, but because they have done the work that mainnet implies.

If a gate is missing, the skill names what is missing and routes the user to the relevant skill. The user can override and proceed by typing the literal phrase the skill asks for, but the override is logged in deploy-context.md so it is on the record.

When to use it

• After the testnet deploy was exercised by a frontend or PTB and the relevant anti-slop skills have run.
• For a production launch the user is willing to defend.

When NOT to use it

• Pre-MVP, before there is a working testnet deploy.
• For experimental projects that should stay on testnet. Mainnet introduces real cost and real user expectations.
• If the user only wants a "label" of mainnet without the responsibility. Route to validate-business-model and retention-loop first.

If you activated this and the user actually wants something else, consult skills/SKILL_ROUTER.md and hand off.

Inputs

• .suiperpower/business-model.md (required, verdict must be yes or partial).
• .suiperpower/retention-loop.md (required, verdict must not be no).
• A recent review-move output, indicated by an entry in .suiperpower/learnings.md under Decisions or a separate .suiperpower/review-move.md if present.
• .suiperpower/deploy-context.md (required, must contain at least one testnet entry).
• Move.toml and the package path for the publish.
• A funded mainnet address.

Outputs

A ## Deploy, <timestamp> block appended to .suiperpower/deploy-context.md with env: mainnet, the package id, the upgrade cap, the upgrade policy chosen, and any override notes:

## Deploy, <timestamp>

### Network
- env: mainnet
- rpc: <rpc url>
- gas address: <0x...>
- balance before: <SUI>
- balance after: <SUI>

### Package
- package id: <0x...>
- modules: <comma-separated module names>
- digest: <publish digest>
- upgrade cap object: <0x...>
- upgrade cap policy: <compatible | additive | dep-only | immutable>

### Gates
- business-model: <yes | partial | overridden>
- retention-loop: <yes | partial | overridden>
- review-move: <recent | overridden>
- testnet exercised: <yes | overridden>

### Notes
- <override rationale, if any>
- <stakeholders notified, if any>

Workflow

1. Read the gate inputs

• Open business-model.md, retention-loop.md, deploy-context.md. If any are missing, surface which ones and route the user back.
• Look for evidence that review-move ran in the last 14 days. The skill considers a ## Decisions entry referencing review-move or a .suiperpower/review-move.md file as evidence.

1. Score the gates

• business-model: yes to proceed cleanly, partial to proceed with a flag, no to refuse unless override.
• retention-loop: yes or partial to proceed, no to refuse unless override.
• review-move: recent run within 14 days, or refuse unless override.
• testnet exercised: must have a testnet entry in deploy-context.md plus user-confirmed evidence the testnet package was exercised by a frontend or PTB.

1. If a gate fails

• Name the failed gate explicitly.
• Recommend the corresponding skill to clear it: validate-business-model, retention-loop, review-move, deploy-to-testnet.
• Offer the override path: the user must type "I accept the risk and want to deploy anyway" verbatim. Anything less is treated as a no.
• If the user overrides, record the rationale in the Notes section of the new deploy entry.

1. Mainnet preflight

• Verify Sui CLI version matches the version pinned in skills/data/guides/deploy-runbook.md.
• Verify env is mainnet, address is funded, RPC is healthy.
• Verify the same Move package builds and tests pass: sui move build, sui move test.

1. Publish

• Run sui client publish <package path> --gas-budget <budget> --json.
• Capture JSON output to a temp file.

1. Capture artifacts

• Apply jq recipe per skills/data/guides/package-id-capture.md.
• Extract package id, digest, UpgradeCap id.

1. Lock the upgrade policy intentionally

• Default policy after first mainnet publish is compatible. The skill prompts the user to choose:
• compatible (default): future upgrades may add fields and functions but not change existing signatures.
• additive: future upgrades may only add, never modify.
• dep-only: future upgrades only update dependency package addresses.
• immutable: no future upgrades possible. Most secure, most rigid.
• The skill records the chosen policy. If the user picks immutable, the skill confirms with a second prompt.

1. Write deploy-context.md

• Append the new mainnet block with all artifacts and the gate status.
• Leave the testnet entry intact.

1. Verify on-chain

• sui client object <package id> and sui client object <upgrade cap id>.
• Surface the package id to the user with explicit "this is mainnet" framing.

1. Hand off

• Recommend submit-to-sui-overflow if this is a hackathon submission.
• Recommend apply-grant if the user is preparing a Sui Foundation grant application.
• Recommend analytics-baseline (post-launch grow phase) once it ships in v1.1.

Quality gate (anti-slop)

Before reporting done:

• Were all four gates evaluated explicitly, with a recorded status?
• Was the override path required to be the verbatim phrase, not a casual "yes"?
• Was the upgrade policy selected intentionally (not silently defaulted)?
• Was the new entry appended to deploy-context.md with env: mainnet and all required fields?
• Was the post-deploy sui client object confirmation surfaced?

If any answer is no, the skill keeps working.

References

On-demand references (load when relevant to the user's question):

• references/gate-rationale.md: Why each gate exists and what failure looks like.
• references/upgrade-policy-choice.md: Trade-offs between compatible, additive, dep-only, immutable.

Canonical:

• skills/data/guides/deploy-runbook.md
• skills/data/guides/package-id-capture.md
• skills/data/guides/security-checklist.md

Use in your agent

• Claude Code: claude "/suiper:deploy-to-mainnet <your message>"
• Codex: codex "/deploy-to-mainnet <your message>"
• Grok Build: run grok, then /deploy-to-mainnet <your message> in the session
• Cursor: paste a chat message that includes a phrase like "deploy to mainnet", or load ~/.cursor/rules/deploy-to-mainnet.mdc and reference it.

If you activated this and the user actually wants something else, consult skills/SKILL_ROUTER.md and hand off.