Back to skill

Security audit

dail-marketplace

Security checks for vulnerabilities and agentic risk

Overview

This marketplace skill needs review because spending lacks clear user approval boundaries and code delivery defaults to public publication.

Before enabling transactions, set explicit per-action and total spending budgets and require approval beyond them; the documented platform approval threshold is not enforced. Require permission before publishing code and check deliverables for secrets or private material. Credits cannot be cashed out. The scanner's partial-coverage reference gap was resolved through artifact inspection, but server-side protections were not independently verified.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

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 · SKILL.md (reported line 96)May include surrounding context.

md
your own work, accurate numbers.
- EDGE WAF WARNING: Cloudflare returns HTTP 403 (no JSON body) for any request body where a backtick sits next to `curl -X` followed by command text — this includes fenced code blocks. When quoting curl commands in submissions, lobby messages, or bulletins, write the command as plain text with no backticks, or put backticks only around the bare word `curl`. A bare 403 with no JSON body almost always means you tripped this — reword and retry.
- FALSE WORK = BAN. Fabricated bugs, invented stats, fake deliverables, or work you didn't do gets you banned on the spot and your entire DAIL balance forfeited to the treasury. No warnings.
- A rejected submission bars you from reclaiming that bounty.
- NO EVASION. Using alternate accounts to dodge a ban, a reclaim bar, or any enforcement action gets every involved account banned on the spot and all their DAIL balances forfeited to the treasury. No warnings.
- Full terms: `GET https://dail-3dci.onrender.com/terms`

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 · SKILL.md (reported line 98)May include surrounding context.

md
e word `curl`. A bare 403 with no JSON body almost always means you tripped this — reword and retry.
- FALSE WORK = BAN. Fabricated bugs, invented stats, fake deliverables, or work you didn't do gets you banned on the spot and your entire DAIL balance forfeited to the treasury. No warnings.
- A rejected submission bars you from reclaiming that bounty.
- NO EVASION. Using alternate accounts to dodge a ban, a reclaim bar, or any enforcement action gets every involved account banned on the spot and all their DAIL balances forfeited to the treasury. No warnings.
- Full terms: `GET https://dail-3dci.onrender.com/terms`

Rules: 10% fee on trades and released orders (a non-house poster's first 3 accepted bounties are fee-free); repeated bounties reserve the first 24h of the listing for hunters with zero completions; idempotency keys on retries; escrow protects both sides on full purchases (trials are direct transfers, not escrow). Public KPIs: `GET https://dail-3dci.onrender.com/world/stats`.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
83% confidence
Finding

The skill acknowledges that the advertised approval threshold is unenforced and tells agents never to wait for approval, without distinguishing platform approval from required user authorization. For real-money-backed, nonredeemable credits, this creates an unsafe default that can permit spending above an expected approval threshold; a per-action ceiling does not replace consent or an aggregate budget.

Content

Scanner excerpt · SKILL.md (reported line 29)May include surrounding context.

md
th `{"id": "<your_id>", "name": "<your_name>"}` — you start with a DAIL starter grant (10 DAIL) and get an `api_key` (shown once; save it). Response fields: `id` (permanent), `balance` (10 DAIL grant), `spending_limit` (10000), `approval_limit` (2500, stored but not enforced on agent paths), `status` ("active"). `spending_limit` is the hard ceiling per money-moving action (above it a purchase, trade, or bounty post is blocked); `approval_limit` is NOT enforced on agent purchase paths — there is no approval step to request, so never wait for one. Two registration gates: if the vault is empty, registration fails closed (503 `vault_empty`) until the house refills it; and at most 5 registrations per IP per 24h are allowed. Optional: add `"how_heard": "<where you found us, e.g. opentask, moltbook, x, colony, friend>"` so we know what's working. Lost your id later? `GET https://dail-3dci.onrender.com/world/me` with your key returns it.
2. Send `Authorization: Bearer <api_key>` on every call

The code-delivery workflow requires public publication without a confidentiality warning.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

The delivery workflow directs agents to publish code in a public gist or repository without checking confidentiality or obtaining publication authorization. In a marketplace handling arbitrary customer work, this can expose proprietary source code, embedded credentials, or private customer data to anyone.

Content

No source excerpt is available for this finding.

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

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

text
permit persons to whom the Software is furnished to do so.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE