FC rustok-wallet-tui
Self-custody Ethereum agent wallet. Installs from a verified release and runs entirely on your machine as a single container image (MCP over stdio); private keys never leave it. New connections are read-only. Read wallet context, balances and DeFi positions (Aave v3, ERC-4626); preview transactions. Sending funds on-chain is gated in a separate terminal console, never inside the agent chat: you approve each payment, or confirm autonomous mode once and the wallet sends on its own after that; message signing is refused outright, in every mode. You assume all risk for funds on the agent wallet — there are no hard-coded spending limits.
Self-custody Ethereum agent wallet.
As a process C 62/100 · Has gaps — weak spots: result and completion, when it triggers, consistency
What is at stake
Medium-severity findings: the skill is probably honest, but read what alarmed the scanner.
Below is the worst case for this category. The finding here is medium: the guard saw a sign, not a proof.
The skill contains commands that delete files, rewrite disks or execute code fetched from the network. The agent may run them without asking if it believes the instructions require it.
Replace destructive commands with safe equivalents that ask for confirmation, scope them to one folder, and stop piping curl into a shell: pin a version and a checksum.
Below is the worst case for this category. The finding here is medium: the guard saw a sign, not a proof.
The purpose itself is risky: wallets, browser password stores, offensive security. Even an honest implementation gives the agent access to things that cost money.
Explain in the description why the access is needed and how it is limited; add tests that show refusals on dangerous requests.
How to improve
- Say in the description WHEN to use the skill ("use when…", example requests): that is the agent's main cue.
- The SKILL.md body is over 5,000 tokens: move reference detail into references/ and load it when needed.
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 12
✓ No critical or high findings
Medium and low: 12
-
medium Dangerous commands
cmd-shell-rcinstall.sh:90Writes to a shell startup filebash) echo "$HOME/.bashrc" ;;
-
medium Dangerous commands
cmd-shell-rcinstall.sh:91Writes to a shell startup filezsh) echo "$HOME/.zshrc" ;;
-
medium Dangerous commands
cmd-shell-rcinstall.sh:92Writes to a shell startup filesh | ksh | dash | ash) echo "$HOME/.profile" ;;
-
medium Risky intent
intent-wallet-secretsskill-card.md:29Handles crypto-wallet secrets (seed / mnemonic / private key) — a classic stealer targetRisk: A leaked approval PIN or recovery phrase can compromise wallet control or recovery.
-
medium Risky intent
intent-wallet-secretsskill-card.md:31Handles crypto-wallet secrets (seed / mnemonic / private key) — a classic stealer targetMitigation: Enter the PIN and recovery phrase only in your own terminal; never paste either into agent chat, and store the recovery phrase offline.
-
medium Risky intent
intent-wallet-secretsSKILL.md:173Handles crypto-wallet secrets (seed / mnemonic / private key) — a classic stealer targetexactly one thing, once: the **12-word recovery phrase**.
-
medium Risky intent
intent-wallet-secretsSKILL.md:199Handles crypto-wallet secrets (seed / mnemonic / private key) — a classic stealer targethave their recovery phrase backed up offline before proceeding; never ask them
-
medium Risky intent
intent-wallet-secretsSKILL.md:210Handles crypto-wallet secrets (seed / mnemonic / private key) — a classic stealer target`delete my keys` confirmation before changes. Without the recovery phrase,
-
low Risky intent
intent-wallet-secretsinstall.sh:211Handles crypto-wallet secrets (seed / mnemonic / private key) — a classic stealer target (quoted — discussed, not commanded)say "next in your own terminal: rustok init — you choose a 6-digit PIN, and see your recovery phrase once"
quoted -
low Obfuscation
obf-base64-blobrelease.manifest.sigstore.json:1Long base64-looking blob (quoted — discussed, not commanded){"mediaType":"application/vnd.…0.3+json","verificationMaterial":{"certificate":{"rawBytes":"MIIH…Pr9+d5p4…3Rvquoted -
low Secrets in code
secret-high-entropy-tokenrelease.manifest.sigstore.json:1High-entropy token-like string (may be an id, hash or a credential) (quoted — discussed, not commanded){"mediaType":"application/vnd.…0.3+json","verificationMaterial":{"certificate":{"rawBytes":"MIIH…Pr9+d5p4…3Rvquoted -
low Risky intent
intent-wallet-secretsSKILL.md:39Handles crypto-wallet secrets (seed / mnemonic / private key) — a classic stealer target (detector / deny-list definition; documentation table row)| **Outside the model** | An agent with **shell / `docker exec` access to the container** can read secrets and bypass client isolation. Direct message and typed-data signing are refused by the core in
detectortable
Files scanned: 5. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- warning
frontmatter-yamlSKILL.md: the frontmatter is not valid YAML (YAML parse error: Nested mappings are not allowed in compact mappings at line 2, column 14: description: Self-custody Ethereum agent wallet. Installs from a verified relea… ^ ); fields were read line by line. The usual cause is a colon inside an unquoted value - warning
description-no-whendescription does not say WHEN to use the skill (no "use when") - warning
body-longSKILL.md body ≈ 5573 tokens (recommended < 5000); move details to references/
Process rating: all ten parameters 62/100
- 20When it triggers. No condition that starts the skill
- 30Running it twice. 20 mutating operations with no state check
- 40Result and completion. Does not say what the result is
- 40Consistency. Frontmatter name (rustok-wallet-tui) differs from the folder (wallet)
- 60Tools and files. Uses tools (bash, web, node) that frontmatter does not declare
- 70Inputs and preconditions. Inputs and preconditions are listed
- 70Execution cost. Instruction body is 5573 tokens
- 100Steps. 15 steps
- 100Failures and branches. 2 branches, has a failure section
- 100Progress reporting. Reports progress
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
- low 13 top-level sections: this looks like several domains in one skill
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +5Description has no quoted example phrases that should trigger the skill
- +4Description does not say when NOT to use the skill (false activations)
- +3Output format is not stated: the model decides each time
- +1No license
- +2Single-language instructions
- +3Description length 640: enough signal without eating the budget
- +4Structure: 16 headings
- +3Step-by-step instructions: 15 items
- +4Has examples (4 code blocks)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 52.