Back to skill

Security audit

Pursekeeper

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious because scanners report a network redirection weakness and a payment-lock race, while inspection was incomplete; these results do not establish malicious intent.

Consider the risk of requests reaching internal services and the reported uncertainty around payment outcomes before installing. Review the external services and data-sharing instructions, and note that the scan covered only 44.4% of components fully, so its evidence is incomplete.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/client-x402.js
Finding

Seller-controlled work redirects reach unauthorized network destinations

Content
View full analysis

Impact

Blind SSRF permits requests to internal or local HTTP services from the client's network position. Request bodies are constrained; this path does not transmit the seed, caller HEADERS, or a signed payment block.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
Findings (26)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 12)May include surrounding context.

md
description: 64 hex characters. The wallet seed read by scripts/no-node.js and scripts/client-x402.js. Make one with `openssl rand -hex 32`, store it with mode

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 49)May include surrounding context.

md
`scripts/no-node.js` is a complete wallet in one file.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 15)May include surrounding context.

md
description: Spending cap for scripts/client-x402.js in NANO as decimal text (0.05, 1, 0.000001), default 0.01. A 402 quote above it is refused and nothing is s

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 18)May include surrounding context.

md
description: Optional path. After a scripts/client-x402.js call, the PAYMENT-SIGNATURE header value is written there (mode 600) with the url, block hash and fin

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 21)May include surrounding context.

md
description: scripts/client-x402.js holds a per-account lock on this host from the frontier read until the fate of its block is known, so several processes payi

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 121)May include surrounding context.

md
NANO_SEED=... node scripts/client-x402.js https://pursekeeper.dev/v1/echo?msg=hello

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 122)May include surrounding context.

md
NANO_SEED=... NANO_MAX_PAY=0.05 node scripts/client-x402.js https://some-seller.example/paid # cap in NANO; default 0.01, a higher quote is refused unsigned

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 123)May include surrounding context.

md
NANO_SEED=... METHOD=POST BODY='{"records":[...]}' node scripts/client-x402.js https://nano-csv-service.onrender.com/clean

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 124)May include surrounding context.

md
NANO_SEED=... NANO_SAVE_PAYLOAD=./proof.json node scripts/client-x402.js https://seller.example/buy # keeps the payment header for a later download route; mode

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 21)May include surrounding context.

md
holder's record runs under an exclusive `.reaper` file beside the lock, scripts/wallet-lock-reaper.js, so two waiters cannot both reclaim one record, and a `.re

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 24)May include surrounding context.

md
description: Cumulative spending cap for scripts/inference-proxy.js in NANO as decimal text, over the life of its state file. The proxy refuses to start without

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 27)May include surrounding context.

md
description: The one door scripts/inference-proxy.js may pay, e.g. https://nano-gpt.com/api/x402. Quotes whose status or complete URLs sit on another host are r

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · SKILL.md (reported line 109)May include surrounding context.

change daily, and each live entry carries a reachability probe.

bash
curl -s https://pursekeeper.dev/sellers.json | node -e '
  let d="";process.stdin.on("data",c=>d+=c).on("end",()=>{for(const s of JSON.parse(d).sellers)
  console.log(s.live.reachable?"up  ":"down",s.id,"|",s.price,"|",s.endpoint)})'

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
80% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · scripts/no-node.js (reported line 145)May include surrounding context.

js
const msg = String(e.message);
        console.error('process failed (' + msg.slice(0, 80) + '); checking whether the block landed');
        // The follow-up read must not be allowed to fail on its own: unprotected, its account_info error ended the command
        // with no warning naming the hash, and a rerun of `send` built on the landed block and paid twice (Ops Control HQ,
        // 2026-09-28 22:28 UTC). Same rule as landed(): not knowing is not "did not land".
        try { await refresh(); }
        catch (e2) { throw stop('process failed (' + msg.slice(0, 80) + ') and the follow-up account_info read failed (' + e2.message + '): cannot tell whether ' + hash + ' landed'); }

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · scripts/wallet-lock-reaper.js (reported line 5)May include surrounding context.

js
const fs = require('node:fs');

// Same-host, cooperative wallet clients only. The old compare/unlink sequence
// was not atomic: waiter C could validate dead holder H, waiter B could remove
// H and publish B's *live* record, and C could then delete B's new record.
// Serialise the WHOLE stale-holder recheck + unlink with an O_EXCL local file.
//

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · test/inference-proxy.js (reported line 134)May include surrounding context.

js
}

async function startProxy(env, dir) {
  const proc = spawn(process.execPath, [SCRIPT], { env: { PATH: process.env.PATH, HOME: dir, ...env }, stdio: ['ignore', 'pipe', 'pipe'] });
  let err = '';
  proc.stderr.on('data', c => { err += c; });
  const port = await new Promise((res, rej) => {

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · test/inference-proxy.js (reported line 486)May include surrounding context.

js
{ NANO_SEED: seed, UPSTREAM: 'https://example.invalid', NANO_CAP: '0.1', PROXY_KEY: '' },
    { NANO_SEED: seed, UPSTREAM: 'https://example.invalid', NANO_CAP: '0.1', PROXY_KEY: 'short-key' },
  ]) {
    const code = await new Promise(r => spawn(process.execPath, [SCRIPT], { env: { PATH: process.env.PATH, HOME: dir, STATE_FILE: join(dir, 's.json'), PROXY_KEY: KEY, ...env }, stdio: 'ignore' }).on('exit', r));
    assert.equal(code, 2, JSON.stringify(env));
  }
});

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/buy-from-nanogpt.md (reported line 10)May include surrounding context.

1. Ask, get a 402 with a price and a deposit address

text
curl -s -D headers.txt -o quote.json -X POST https://nano-gpt.com/api/x402/v1/chat/completions \
  -H "Content-Type: application/json" -H "x-x402: nano" --data-binary @request.json

The workflow sends chat messages to NanoGPT without a privacy warning.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The POST uploads request.json, which contains the chat messages, to the third-party NanoGPT service; step 3 uploads the same content again. Although the destination is shown, the instructions do not warn that potentially sensitive conversation content is disclosed to an external provider.

Content

No source excerpt is available for this finding.

Unbounded Resource Access

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill allows unbounded resource consumption (API calls, storage, compute). Without rate limits or quotas, a compromised or misbehaving agent can cause denial-of-service or cost overruns.

Content

Scanner excerpt · references/inference-proxy.md (reported line 214)May include surrounding context.

base_url = "http://127.0.0.1:3402/v1" env_key = "PROXY_KEY" wire_api = "responses" request_max_retries = 0 stream_max_retries = 0

text

Unbounded Resource Access

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill allows unbounded resource consumption (API calls, storage, compute). Without rate limits or quotas, a compromised or misbehaving agent can cause denial-of-service or cost overruns.

Content

Scanner excerpt · references/inference-proxy.md (reported line 215)May include surrounding context.

env_key = "PROXY_KEY" wire_api = "responses" request_max_retries = 0 stream_max_retries = 0

text

`env_key` names the proxy key as the bearer the runtime presents, the one credential the proxy

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/sellers.md (reported line 87)May include surrounding context.

md
estation of any JSON)
- id: `oreomuncher-attest`; operator: OreoMuncher45 (agent operator; GitHub OreoMuncher45), api.shehriyar.ink, 38 pay-per-call endpoints
- what: Signs a canonical (sorted-key) JSON payload with the service's Ed25519 key and returns the canonical sha256, raw sha256, signature and public key; optional label and Base block-number anchor. Independently verifiable: recompute the canonical hash, verify the signature over the sorted-key JSON of the attestation object.
- endpoint: https://api.shehriyar.ink/v1/attest/response
- price: Ӿ0.01 per call (or $0.005 USDC on Base; both rails in one accepts array)
- how to pay: Stock x402 v2: 402 with PAYMENT-REQUIRED carrying two accepts entries; the nano:mainnet one is scheme exact, XNO, 0.01 in raw, payTo nano_1zqdw3…, maxTimeoutSeconds 300, extra work required at fffffff800000000. Sign a send block and retry with PAYMENT-SIGNATURE; the seller settles through facilitator.pursekeeper.dev and answers 200 with PAYMENT-RESPONSE. ap

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/sellers.md (reported line 90)May include surrounding context.

md
RED carrying two accepts entries; the nano:mainnet one is scheme exact, XNO, 0.01 in raw, payTo nano_1zqdw3…, maxTimeoutSeconds 300, extra work required at fffffff800000000. Sign a send block and retry with PAYMENT-SIGNATURE; the seller settles through facilitator.pursekeeper.dev and answers 200 with PAYMENT-RESPONSE. api/examples/client-x402.js works unchanged. Python x402ResourceServer with x402-nano-exact installed from a durable path. Only this one endpoint has the Nano rail so far.
- docs: https://api.shehriyar.ink/openapi.json
- verified 2026-09-11 by block `28E17A7ECBDCCCD10A4A9E21DE541514F255135B24256FC567FBC727A4CA154C` (two Ӿ0.01 calls from my x402 test account (blocks 28E17A7E… and BAC1F9DE…): 200 in 32 s and 15 s (the difference is my own work generation), both settled at facilitator.pursekeeper.dev/stats; canonical hash recomputed and the Ed25519 signature verified locally over the sorted-key attestation JSON)
- note: First existing production x402 API (USDC on Base, Solva

The wallet lock is released even when the signed payment's fate remains unknown, contradicting the documented locking guarantee.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

When reconciliation cannot determine the signed block's fate, explainFate returns null, but the process still exits and its exit handler releases the wallet lock. A subsequent client can sign against the same frontier while the seller still holds a spendable earlier block, enabling competing payments, delayed unexpected settlement, and failed purchases; the spending cap limits individual payments but does not resolve this race.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
93% confidence
Finding

The caret range permits fresh installations to resolve newer compatible releases rather than a single reviewed artifact; no lockfile is supplied in this source. A compromised eligible release could execute with access to the wallet seed and signing operations, making this supply-chain exposure particularly consequential for a cryptocurrency wallet, although no actual dependency compromise is demonstrated.

Content

Scanner excerpt · scripts/package.json (reported line 5)May include surrounding context.

json
"name": "pursekeeper-skill-scripts",
  "private": true,
  "description": "Scripts for the pursekeeper ClawHub skill: a Nano wallet with no node, and an x402 exact client for nano:mainnet.",
  "dependencies": { "nanocurrency": "^2.5.0" }
}