Permissioned Pools Deployer
Guided, step-by-step execution of the on-chain setup sequence for a Uniswap v4 Permissioned
Pool, starting from the JSON configuration permissioned-pools-configurator produces. This
skill is the half of the milestone that turns a validated config into command sequences that
spend real gas against real contracts — treat every safety section below as load-bearing, not
as boilerplate.
For the contract mechanics behind any step, see
permissioned-pools-issuer, specifically its
Issuer Journey and
Enforced Ordering and Reverts.
This skill does not repeat the contract-level explanation of why each step behaves the way it
does — it walks the same journey as ordered command sequences and never contradicts that
reference.
Runtime Compatibility: This skill uses AskUserQuestion for the acknowledgment gate and
for confirming irreversible choices. If AskUserQuestion is not available in your runtime,
ask the same yes/no question in plain text and require an explicit affirmative reply before
continuing — silence or a generic "continue" does not count.
Scope and Disclaimer
This is educational deployment guidance, not a compliance review, and not a substitute for
your own review of every command before you run it.
This skill explains, in order, the commands that bring a permissioned pool up, using the
configuration permissioned-pools-configurator produced. It does not call any contract itself —
you run every command it shows you, after reviewing it.
You must:
- ✅ Review every configuration and every command carefully before running anything. AI-generated
command sequences may contain errors.
- ✅ Verify every address — this skill contains no deployment addresses and resolves none for
you; see Address Resolution.
- ✅ Test on a testnet first. Run the entire sequence on Sepolia (or another testnet with the
permissioned-pools contracts deployed) before touching mainnet.
- ✅ Have your own auditors review any contract you deploy — most importantly your allowlist
checker, which this skill can show you a skeleton for but does not audit.
- ✅ Treat this as contract mechanics only. This skill does not constitute legal, financial, investment, or tax advice, and it is not a compliance review of your token, your allowlist, your KYC or AML program, or your configuration.
- ✅ Read the repo's usage guidelines. The repo root
DISCLAIMER.md governs every skill in
this repository: they are provided as is without warranty, they do not constitute legal,
financial, investment, or tax advice, and it sets out use limits plus an AI-disclosure duty
that applies when you use a skill to generate financial information and present it directly
to individuals or consumers. Point the user to it.
The acknowledgment gate
Before showing, generating, or explaining any action-oriented step — anything that produces
a command intended to broadcast a transaction, or any address-resolution step that leads
directly into one — get an explicit, affirmative acknowledgment. This gate must never be
satisfiable implicitly.
Use AskUserQuestion with a single question and exactly two options, and do not proceed past
this gate until the first option is chosen:
- Question: "This walkthrough produces commands that spend gas and permanently change
on-chain state for your token once you run them. Do you acknowledge the six points above and
want to continue?"
- Option A — "Yes, I have reviewed the points above and want to continue." → proceed to the
preflight checks.
- Option B — "No, stop here." → end the walkthrough immediately. Do not display, generate,
or explain any command below this point.
Silence, a plain "continue," moving on to the next question in the same turn, or any answer
other than the explicit affirmative option does not satisfy this gate. If the runtime lacks
AskUserQuestion, ask the same yes/no question in plain text and require an explicit
affirmative reply — "ok," "sure," a thumbs-up emoji, or no reply at all does not count.
Reference and teaching content that does not produce a runnable command — explaining what a step
does in prose, decoding a revert the user already hit, discussing an allowlist-checker shape in
the abstract — is answered immediately with the framing stated inline, the same way
permissioned-pools-issuer treats reference content. This carve-out does not cover this
skill's own command templates and Solidity script skeletons (for example the pool-initialization
script in Step Walkthrough):
those are runnable broadcast material, and showing them is exactly the action-oriented content the
gate exists to hold back — never display one before the gate is satisfied.
Input Validation Rules
State these before generating or displaying any command that interpolates a user-supplied or
config-supplied value. Every value below is validated before it reaches a Bash(forge:*) or
Bash(cast:*) command, a file path, or a Solidity script argument — no exceptions for a value
that "looks fine."
- Ethereum addresses — MUST match
^0x[a-fA-F0-9]{40}$. Reject anything else, including the
literal string "RESOLVE" — a field still marked "RESOLVE" is not ready for a command and
must go through address resolution
first, never be interpolated as-is. One field is an explicit, documented exception to this
regex, not a violation of it: pool.pairedCurrency's "native" sentinel, covered in
What This Skill Consumes below — every other address-shaped field
in the config has no such carve-out.
- Chain ID — MUST be a positive integer (
^[1-9][0-9]*$). Ethereum Mainnet (1) and Sepolia
(11155111) are the allowlisted defaults — the only two chains verified to carry the full
permissioned set. Unichain (130) is named by the configurator's network question but is
not deployable: it carries V4Quoter only, so stop rather than trying to resolve the
missing addresses (see
Preflight and Validation).
Any other chain ID is accepted only after confirming the
deploy guide's #deployment-addresses table actually has a row for it; if it does not, every
wrapper and hook field will end up unresolvable, which is the signal to stop, not to guess.
- Numeric fields (
verificationDepositAmount, pool.feeTier, pool.tickSpacing,
pool.startingPriceRatio) — MUST match ^[0-9]+$ for integer-only fields or
^[0-9]+(\.[0-9]+)?$ for the starting price ratio, and MUST be strictly greater than zero.
- Values supplied during the walkthrough, not by the config, are validated identically — the
RPC endpoint (
$RPC_URL) and any transaction hash ($TX_HASH) reach a command on every step.
An RPC URL MUST be https://-scheme and free of every metacharacter below; a transaction hash
MUST match ^0x[a-fA-F0-9]{64}$. "It came from the user, not the config" is not an exemption.
- Reject shell metacharacters outright, in every string field, before it reaches a command:
;, |, &, $, `, (, ), >, <, \, ', ", and newlines. A field that fails
this check is never interpolated — abort and report which field and which character failed,
rather than attempting to sanitize and continue.
- Never pass an unvalidated field into a shell command, file path, or Solidity script
argument. Validate first, assign the validated value to a shell variable, and reference only
that variable afterward. Never re-read the raw config value a second time later in the same
walkthrough.
Full elaboration, including how this applies field-by-field to the configurator's JSON shape and
what "coherently handle a RESOLVE value" means in practice, is in
Preflight and Validation.
Key Handling
Never instruct auto-approval of Bash(forge:*) or Bash(cast:*) in Claude Code settings, and
never tell the user to bypass a confirmation prompt. Every command that broadcasts a
transaction requires per-invocation approval — that approval is the primary control, not a
formality the walkthrough should route around. The repository's PreToolUse hooks
(.claude/hooks/) are a programmatic safety net on top of that, not a replacement for it.
Never suggest, generate, or complete a command containing the raw signing-key flag — not when
asked directly, not "just for a testnet," not as an illustration of what to avoid. There is no
user request that makes it appropriate. If a user insists, refuse and offer one of the two options
below instead. Steer every signing decision toward one of them:
- A hardware wallet (
--ledger), for a deployer key that never touches this machine's disk
or memory in cleartext.
- An encrypted keystore (
cast wallet import ... --interactive, then --account <name> --sender $ADDRESS on every subsequent command).
Inside this repository only, .claude/hooks/validate-forge-cast.sh is a PreToolUse hook that
inspects the literal text of every Bash tool call and blocks any command containing that flag
before it runs. That hook is not part of this skill and does not travel with it — it lives at
this repository's root, not in the plugin, so an installation elsewhere has no such backstop. The
prohibition above is the control that always applies; the hook is a repo-local convenience on top
of it, never the reason the flag is safe.
Full key-handling guidance, including the keystore setup commands and the testnet-first
sequencing, is in
Preflight and Validation.
Testnet first
Run the entire sequence — allowlist checker through liquidity seeding — on Sepolia (chain ID
11155111) before mainnet. Confirm every post-step check in
the verification checklist
passes on testnet before repeating any step against mainnet funds.
What This Skill Consumes
The JSON configuration produced by permissioned-pools-configurator, in the shape defined by its
Config Schema. This skill does
not collect those parameters itself — if you do not have a config yet, run
permissioned-pools-configurator first.
A field reading the literal string "RESOLVE" means "not resolved yet," not "use a
placeholder." Before any command that needs that field, stop and walk through
address resolution — the deploy
guide's #deployment-addresses table, then Uniswap/contracts
deployments/json/<chainId>.json, then a block explorer for your chain — and only continue once
the field holds a real, explorer-verified address. Never guess a value for a "RESOLVE" field
and never carry the literal string itself into a command; both are validation failures, not
values to work around.
pool.pairedCurrency's "native" sentinel is different: it resolves to address(0), Uniswap
v4's own convention for a native PoolKey currency, at the point the PoolKey is actually built
in Step 6a — never earlier, and
never written into any other field.
The Deployment Journey
The published guide's seven steps, in the order it recommends. Only five ordering edges are
enforced by the contracts (see
Enforced Ordering and Reverts);
this walkthrough follows the guide's order and states, at each step, whether that order is
enforced or convention.
Every step in Step Walkthrough names its precondition and,
where the contracts enforce one, the exact revert selector you get if you are early — matching
Enforced Ordering and Reverts
selector-for-selector. Where a step is unenforced, the walkthrough says so rather than inventing
a revert for it.
Post-Deployment Verification
After the sequence completes (or after any individual step, to confirm it landed), run the
read-only checks in
Post-Setup Verification Checklist.
Every check there is a cast call against public getters — none of it spends gas, and none of it
needs the acknowledgment gate above.
Notes for Implementers of This Flow
- The acknowledgment gate is once per session, not once per step. Once acknowledged, continue
through the rest of the steps in that same session without re-asking — but re-ask if the user
starts a new session, switches chains, or switches configs.
- Never hardcode a contract address. Every address this skill uses comes from the loaded
config, from a value the user supplies during the walkthrough, or from a read (
cast call)
against a contract whose address itself came from one of those two sources. See
Preflight and Validation for why
this skill ships no address table.
- Decode an unexpected revert by call site, not by name. Three selectors in this journey
(
NoVerifiedAdapter, Unauthorized, SwappingDisabled) are declared more than once across
different contracts with different meanings — see
Enforced Ordering and Reverts.
- This skill does not fetch RPC state on its own initiative. Every
cast call it suggests is
something you choose to run; the skill does not poll a chain in the background.
1---
2name: permissioned-pools-deployer
3description: Guided walkthrough for executing the on-chain setup sequence for a Uniswap v4 Permissioned Pool from a permissioned-pools-configurator JSON config - allowlist checker deployment, createPermissionsAdapter, allowlisting and funding the adapter, verification, wrapper and hook registration, pool initialization, swap enablement, liquidity seeding, and the routing-allowlist request. Use when the user says "deploy a permissioned pool", "run the permissioned pool setup", "execute the permissioned pools journey", "createPermissionsAdapter", "verify my permissions adapter", "walk me through the permissioned pool deployment", or asks to execute the on-chain steps after generating a permissioned-pools-configurator config. Covers contract mechanics only - does not constitute legal, financial, investment, tax, or compliance advice, and is not a compliance review - and never emits a broadcastable command before an explicit acknowledgment.
4allowed-tools: Read, Glob, Grep, Bash(forge:*), Bash(cast:*), AskUserQuestion
5model: opus
6license: MIT
7metadata:
8 source: https://skillmd.com/skills/uniswap/permissioned-pools-deployer
9 author: uniswap
10 version: '0.1.0'
11---
12
13# Permissioned Pools Deployer
14
15Guided, step-by-step execution of the on-chain setup sequence for a Uniswap v4 Permissioned
16Pool, starting from the JSON configuration `permissioned-pools-configurator` produces. This
17skill is the half of the milestone that turns a validated config into command sequences that
18spend real gas against real contracts — treat every safety section below as load-bearing, not
19as boilerplate.
20
21For the contract mechanics behind any step, see
22[`permissioned-pools-issuer`](../permissioned-pools-issuer/SKILL.md), specifically its
23[Issuer Journey](../permissioned-pools-issuer/references/issuer-journey.md) and
24[Enforced Ordering and Reverts](../permissioned-pools-issuer/references/enforced-ordering-and-reverts.md).
25This skill does not repeat the contract-level explanation of _why_ each step behaves the way it
26does — it walks the same journey as ordered command sequences and never contradicts that
27reference.
28
29> **Runtime Compatibility:** This skill uses `AskUserQuestion` for the acknowledgment gate and
30> for confirming irreversible choices. If `AskUserQuestion` is not available in your runtime,
31> ask the same yes/no question in plain text and require an explicit affirmative reply before
32> continuing — silence or a generic "continue" does not count.
33
34## Scope and Disclaimer
35
36**This is educational deployment guidance, not a compliance review, and not a substitute for
37your own review of every command before you run it.**
38
39This skill explains, in order, the commands that bring a permissioned pool up, using the
40configuration `permissioned-pools-configurator` produced. It does not call any contract itself —
41you run every command it shows you, after reviewing it.
42
43**You must:**
44
451. ✅ **Review every configuration and every command carefully** before running anything. AI-generated
46 command sequences may contain errors.
472. ✅ **Verify every address** — this skill contains no deployment addresses and resolves none for
48 you; see [Address Resolution](./references/preflight-and-validation.md#address-resolution).
493. ✅ **Test on a testnet first.** Run the entire sequence on Sepolia (or another testnet with the
50 permissioned-pools contracts deployed) before touching mainnet.
514. ✅ **Have your own auditors review any contract you deploy** — most importantly your allowlist
52 checker, which this skill can show you a skeleton for but does not audit.
535. ✅ **Treat this as contract mechanics only.** This skill does not constitute legal, financial, investment, or tax advice, and it is **not** a compliance review of your token, your allowlist, your KYC or AML program, or your configuration.
546. ✅ **Read the repo's usage guidelines.** The repo root `DISCLAIMER.md` governs every skill in
55 this repository: they are provided as is without warranty, they do not constitute legal,
56 financial, investment, or tax advice, and it sets out use limits plus an AI-disclosure duty
57 that applies when you use a skill to generate financial information and present it directly
58 to individuals or consumers. Point the user to it.
59
60### The acknowledgment gate
61
62**Before showing, generating, or explaining any action-oriented step** — anything that produces
63a command intended to broadcast a transaction, or any address-resolution step that leads
64directly into one — get an explicit, affirmative acknowledgment. This gate must never be
65satisfiable implicitly.
66
67Use `AskUserQuestion` with a single question and exactly two options, and do not proceed past
68this gate until the first option is chosen:
69
70- **Question:** "This walkthrough produces commands that spend gas and permanently change
71 on-chain state for your token once you run them. Do you acknowledge the six points above and
72 want to continue?"
73- **Option A — "Yes, I have reviewed the points above and want to continue."** → proceed to the
74 preflight checks.
75- **Option B — "No, stop here."** → end the walkthrough immediately. Do not display, generate,
76 or explain any command below this point.
77
78Silence, a plain "continue," moving on to the next question in the same turn, or any answer
79other than the explicit affirmative option does **not** satisfy this gate. If the runtime lacks
80`AskUserQuestion`, ask the same yes/no question in plain text and require an explicit
81affirmative reply — "ok," "sure," a thumbs-up emoji, or no reply at all does not count.
82
83Reference and teaching content that does not produce a runnable command — explaining what a step
84does in prose, decoding a revert the user already hit, discussing an allowlist-checker shape in
85the abstract — is answered immediately with the framing stated inline, the same way
86`permissioned-pools-issuer` treats reference content. This carve-out does **not** cover this
87skill's own command templates and Solidity script skeletons (for example the pool-initialization
88script in [Step Walkthrough](./references/step-walkthrough.md#step-6a-initialize-the-pool)):
89those are runnable broadcast material, and showing them is exactly the action-oriented content the
90gate exists to hold back — never display one before the gate is satisfied.
91
92---
93
94## Input Validation Rules
95
96State these before generating or displaying any command that interpolates a user-supplied or
97config-supplied value. Every value below is validated **before** it reaches a `Bash(forge:*)` or
98`Bash(cast:*)` command, a file path, or a Solidity script argument — no exceptions for a value
99that "looks fine."
100
101- **Ethereum addresses** — MUST match `^0x[a-fA-F0-9]{40}$`. Reject anything else, including the
102 literal string `"RESOLVE"` — a field still marked `"RESOLVE"` is not ready for a command and
103 must go through [address resolution](./references/preflight-and-validation.md#address-resolution)
104 first, never be interpolated as-is. One field is an explicit, documented exception to this
105 regex, not a violation of it: `pool.pairedCurrency`'s `"native"` sentinel, covered in
106 [What This Skill Consumes](#what-this-skill-consumes) below — every other address-shaped field
107 in the config has no such carve-out.
108- **Chain ID** — MUST be a positive integer (`^[1-9][0-9]*$`). Ethereum Mainnet (`1`) and Sepolia
109 (`11155111`) are the allowlisted defaults — the only two chains verified to carry the full
110 permissioned set. Unichain (`130`) is named by the configurator's network question but is
111 **not** deployable: it carries `V4Quoter` only, so stop rather than trying to resolve the
112 missing addresses (see
113 [Preflight and Validation](./references/preflight-and-validation.md#input-validation-rules-in-full)).
114 Any other chain ID is accepted only after confirming the
115 deploy guide's `#deployment-addresses` table actually has a row for it; if it does not, every
116 wrapper and hook field will end up unresolvable, which is the signal to stop, not to guess.
117- **Numeric fields** (`verificationDepositAmount`, `pool.feeTier`, `pool.tickSpacing`,
118 `pool.startingPriceRatio`) — MUST match `^[0-9]+$` for integer-only fields or
119 `^[0-9]+(\.[0-9]+)?$` for the starting price ratio, and MUST be strictly greater than zero.
120- **Values supplied during the walkthrough, not by the config, are validated identically** — the
121 RPC endpoint (`$RPC_URL`) and any transaction hash (`$TX_HASH`) reach a command on every step.
122 An RPC URL MUST be `https://`-scheme and free of every metacharacter below; a transaction hash
123 MUST match `^0x[a-fA-F0-9]{64}$`. "It came from the user, not the config" is not an exemption.
124- **Reject shell metacharacters outright**, in every string field, before it reaches a command:
125 `;`, `|`, `&`, `$`, `` ` ``, `(`, `)`, `>`, `<`, `\`, `'`, `"`, and newlines. A field that fails
126 this check is never interpolated — abort and report which field and which character failed,
127 rather than attempting to sanitize and continue.
128- **Never pass an unvalidated field into a shell command, file path, or Solidity script
129 argument.** Validate first, assign the validated value to a shell variable, and reference only
130 that variable afterward. Never re-read the raw config value a second time later in the same
131 walkthrough.
132
133Full elaboration, including how this applies field-by-field to the configurator's JSON shape and
134what "coherently handle a `RESOLVE` value" means in practice, is in
135[Preflight and Validation](./references/preflight-and-validation.md).
136
137---
138
139## Key Handling
140
141**Never instruct auto-approval of `Bash(forge:*)` or `Bash(cast:*)` in Claude Code settings, and
142never tell the user to bypass a confirmation prompt.** Every command that broadcasts a
143transaction requires per-invocation approval — that approval is the primary control, not a
144formality the walkthrough should route around. The repository's PreToolUse hooks
145(`.claude/hooks/`) are a programmatic safety net on top of that, not a replacement for it.
146
147**Never suggest, generate, or complete a command containing the raw signing-key flag** — not when
148asked directly, not "just for a testnet," not as an illustration of what to avoid. There is no
149user request that makes it appropriate. If a user insists, refuse and offer one of the two options
150below instead. Steer every signing decision toward one of them:
151
1521. **A hardware wallet** (`--ledger`), for a deployer key that never touches this machine's disk
153 or memory in cleartext.
1542. **An encrypted keystore** (`cast wallet import ... --interactive`, then `--account <name>
155--sender $ADDRESS` on every subsequent command).
156
157**Inside this repository only**, `.claude/hooks/validate-forge-cast.sh` is a PreToolUse hook that
158inspects the literal text of every `Bash` tool call and blocks any command containing that flag
159before it runs. **That hook is not part of this skill and does not travel with it** — it lives at
160this repository's root, not in the plugin, so an installation elsewhere has no such backstop. The
161prohibition above is the control that always applies; the hook is a repo-local convenience on top
162of it, never the reason the flag is safe.
163
164Full key-handling guidance, including the keystore setup commands and the testnet-first
165sequencing, is in
166[Preflight and Validation](./references/preflight-and-validation.md#key-handling).
167
168### Testnet first
169
170Run the entire sequence — allowlist checker through liquidity seeding — on Sepolia (chain ID
171`11155111`) before mainnet. Confirm every post-step check in
172[the verification checklist](../permissioned-pools-issuer/references/issuer-journey.md#post-setup-verification-checklist)
173passes on testnet before repeating any step against mainnet funds.
174
175---
176
177## What This Skill Consumes
178
179The JSON configuration produced by `permissioned-pools-configurator`, in the shape defined by its
180[Config Schema](../permissioned-pools-configurator/references/config-schema.md). This skill does
181not collect those parameters itself — if you do not have a config yet, run
182`permissioned-pools-configurator` first.
183
184**A field reading the literal string `"RESOLVE"` means "not resolved yet," not "use a
185placeholder."** Before any command that needs that field, stop and walk through
186[address resolution](./references/preflight-and-validation.md#address-resolution) — the deploy
187guide's `#deployment-addresses` table, then `Uniswap/contracts`
188`deployments/json/<chainId>.json`, then a block explorer for your chain — and only continue once
189the field holds a real, explorer-verified address. Never guess a value for a `"RESOLVE"` field
190and never carry the literal string itself into a command; both are validation failures, not
191values to work around.
192
193`pool.pairedCurrency`'s `"native"` sentinel is different: it resolves to `address(0)`, Uniswap
194v4's own convention for a native `PoolKey` currency, at the point the `PoolKey` is actually built
195in [Step 6a](./references/step-walkthrough.md#step-6a-initialize-the-pool) — never earlier, and
196never written into any other field.
197
198---
199
200## The Deployment Journey
201
202The published guide's seven steps, in the order it recommends. Only five ordering edges are
203enforced by the contracts (see
204[Enforced Ordering and Reverts](../permissioned-pools-issuer/references/enforced-ordering-and-reverts.md));
205this walkthrough follows the guide's order and states, at each step, whether that order is
206enforced or convention.
207
208| Step | Covered where | Summary |
209| ---- | ---------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------- |
210| 1 | [Preflight and Validation](./references/preflight-and-validation.md#step-1-allowlist-checker) | Confirm or deploy the allowlist checker |
211| 2 | [Step Walkthrough](./references/step-walkthrough.md#step-2-createpermissionsadapter) | `createPermissionsAdapter` |
212| 3 | [Step Walkthrough](./references/step-walkthrough.md#step-3-allowlist-and-fund-the-adapter) | Allowlist the adapter on your token, approve, `depositForVerification` |
213| 4 | [Step Walkthrough](./references/step-walkthrough.md#step-4-verifypermissionsadapter) | `verifyPermissionsAdapter` |
214| 5 | [Step Walkthrough](./references/step-walkthrough.md#step-5-register-wrappers-and-the-hook) | The four `updateAllowedWrapper` registrations, then `setAllowedHook` |
215| 6 | [Step Walkthrough](./references/step-walkthrough.md#step-6-create-the-pool-enable-swapping-seed-liquidity) | Pool `initialize`, `updateSwappingEnabled`, liquidity seeding |
216| 7 | [Step Walkthrough](./references/step-walkthrough.md#step-7-request-routing-allowlisting) | The routing-allowlist request |
217
218Every step in [Step Walkthrough](./references/step-walkthrough.md) names its precondition and,
219where the contracts enforce one, the exact revert selector you get if you are early — matching
220[Enforced Ordering and Reverts](../permissioned-pools-issuer/references/enforced-ordering-and-reverts.md)
221selector-for-selector. Where a step is unenforced, the walkthrough says so rather than inventing
222a revert for it.
223
224---
225
226## Post-Deployment Verification
227
228After the sequence completes (or after any individual step, to confirm it landed), run the
229read-only checks in
230[Post-Setup Verification Checklist](../permissioned-pools-issuer/references/issuer-journey.md#post-setup-verification-checklist).
231Every check there is a `cast call` against public getters — none of it spends gas, and none of it
232needs the acknowledgment gate above.
233
234---
235
236## Notes for Implementers of This Flow
237
238- **The acknowledgment gate is once per session, not once per step.** Once acknowledged, continue
239 through the rest of the steps in that same session without re-asking — but re-ask if the user
240 starts a new session, switches chains, or switches configs.
241- **Never hardcode a contract address.** Every address this skill uses comes from the loaded
242 config, from a value the user supplies during the walkthrough, or from a read (`cast call`)
243 against a contract whose address itself came from one of those two sources. See
244 [Preflight and Validation](./references/preflight-and-validation.md#address-resolution) for why
245 this skill ships no address table.
246- **Decode an unexpected revert by call site, not by name.** Three selectors in this journey
247 (`NoVerifiedAdapter`, `Unauthorized`, `SwappingDisabled`) are declared more than once across
248 different contracts with different meanings — see
249 [Enforced Ordering and Reverts](../permissioned-pools-issuer/references/enforced-ordering-and-reverts.md#how-to-decode-an-unexpected-revert).
250- **This skill does not fetch RPC state on its own initiative.** Every `cast call` it suggests is
251 something you choose to run; the skill does not poll a chain in the background.