๐Ÿ“ฆ

Audit History

integration-tests - 2 audits

Version comparison

Capability and finding changes across audited versions, newest first.

VersionDateResultReview itemsChange vs previous
v2 LatestJul 21, 2026, 03:20 AM No confirmed findings0No capability change
v1 Jul 10, 2026, 12:00 AM 2 confirmed7Baseline

Jul 21, 2026, 03:20 AM

All 25 external-command findings are false positives caused by Markdown fences and inline code formatting in SKILL.md. The two private-key findings document environment-variable configuration for testnet accounts and show no secret access or exfiltration. No prompt injection, malicious intent, or additional semantic security issue was found.

2
Files scanned
152
Lines analyzed
2
Review items
0
False positives ignored
Audited by: claude

Jul 10, 2026, 12:00 AM

Most external command alerts are Markdown formatting false positives. Confirmed risks remain for gltest setup commands, scoped cache deletion, private-key environment variables, real web or LLM calls, and funded testnet transactions. No evidence of prompt injection was found.

2
Files scanned
152
Lines analyzed
11
Review items
0
False positives ignored

Confirmed security concerns (2)

Medium
Real External Web and LLM Calls in Tests
The skill states integration tests use real web and LLM calls instead of mocks. Test data may leave the local environment or incur provider cost.
The file explicitly contrasts mocks with real web and LLM calls and lists this as an integration-test use case.
Medium
Funded Testnet Transaction Risk
The skill supports transact calls against Testnet Bradbury and notes that funded accounts are required. Running tests can submit real testnet transactions.
The documentation shows .transact() usage and separately identifies Testnet Bradbury as a real network requiring funded accounts.
Capability review items (7)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

High
Generic API/secret keys
- "${ACCOUNT_PRIVATE_KEY_1}"
Line 94 references ACCOUNT_PRIVATE_KEY_1 in a testnet account configuration. Private keys are sensitive secrets, even when read from environment variables.
High
Generic API/secret keys
- "${ACCOUNT_PRIVATE_KEY_2}"
Line 95 references ACCOUNT_PRIVATE_KEY_2 in the same testnet account configuration. This confirms the skill may require handling private keys for funded accounts.
Medium
Ruby/shell backtick execution ยท 2 occurrences
```bash
Lines 17-25 show gltest shell commands that can run tests against local, hosted, or testnet networks. The commands are legitimate but still execute external tooling and may submit transactions.
Medium
Ruby/shell backtick execution
- **GLSim** (`pip install genlayer-test[sim]`, `glsim --port 4000 --validators 5`) โ€” lightweight, no
Line 111 includes pip install and glsim shell commands. Installing packages and starting local services are legitimate setup steps but require user confirmation.
Medium
Ruby/shell backtick execution
- **Studio local** (`genlayer up`) โ€” full GenVM, real consensus, Docker required. Validates runtime
Line 112 instructs running genlayer up, which starts a local Studio environment and may invoke Docker. This is expected but is real external command execution.
Medium
Ruby/shell backtick execution
Clear cache: `rm -rf .gltest_cache`
Line 128 recommends rm -rf .gltest_cache to clear cached state. The path is scoped, but it is still a destructive filesystem command.
Audited by: codex