中文 EN

Scan Report

Scan New Files

⚠ Low Risk — Risk Score 25/100
Last scan:15 hr ago Rescan
25 /100
navimem
Shared web task memory for AI agents. Query community workflow knowledge before browsing and report execution traces after tasks.
NaviMem is a legitimate community memory system for browser task knowledge sharing. All capabilities are declared and aligned with the documented functionality, though privacy implications of external trace sharing warrant awareness.
Skill Namenavimem
Duration57.9s
Enginepi
✓
Safe to install
The skill is safe to use for non-sensitive browser tasks. Be cautious when performing tasks that involve sensitive URLs, tokens, or personal data, as execution traces are transmitted to i.ariseos.com. Verify the credential-stripping claim independently if handling high-sensitivity workflows.

Findings 2 items

Severity Finding Location
Low
Execution traces transmitted to external service Data Exfil
Browser task execution traces (URLs, actions, targets, values) are POSTed to i.ariseos.com. While this is the documented purpose of the skill (community memory), URLs may inadvertently contain sensitive tokens, session IDs, or personal data not caught by generic credential stripping.
-H "Content-Type: application/json" -d '{"type": "browser_workflow", "task": "...", "steps": [...]}
→ Use this skill cautiously for non-sensitive browser tasks. Avoid using it for tasks involving banking, medical, or other high-sensitivity workflows where URL parameters or form data could leak PII.
 SKILL.md:1
Info
Privacy claim lacks technical verification Doc Mismatch
SKILL.md states 'input values and credentials are stripped' but this is a textual claim without evidence of implementation. Generic patterns may miss edge cases (e.g., tokens in URL paths, base64-encoded values).
Privacy: only workflow structure is shared, input values and credentials are stripped
→ Do not rely on automated credential stripping for high-sensitivity data. Manually sanitize traces before submission if using for sensitive tasks.
 SKILL.md:90
ResourceDeclaredInferredStatusEvidence
Filesystem NONE NONE — No file system operations found in SKILL.md
Network READ READ ✓ Aligned SKILL.md declares HTTP POST calls to i.ariseos.com for plan/query/learn endpoint…
Shell NONE NONE — No shell execution in documentation
Environment READ READ ✓ Aligned SKILL.md metadata declares NAVIMEM_BASE_URL env var (optional)
Browser NONE READ ✓ Aligned Core purpose is to assist browser automation tasks (plan/execute/learn loop)
8 findings
🔗
Medium External URL 外部 URL
https://i.ariseos.com/api/v1/memory/plan
README.md:25
🔗
Medium External URL 外部 URL
https://i.ariseos.com/api/v1/memory/learn
README.md:30
🔗
Medium External URL 外部 URL
https://www.amazon.com/
README.md:37
🔗
Medium External URL 外部 URL
https://www.amazon.com/s?k=laptop
README.md:40
🔗
Medium External URL 外部 URL
https://i.ariseos.com
SKILL.md:15
🔗
Medium External URL 外部 URL
https://i.ariseos.com/api/v1/memory/query
SKILL.md:77
🔗
Medium External URL 外部 URL
https://www.amazon.com/cart
SKILL.md:88
🔗
Medium External URL 外部 URL
https://www.amazon.com/checkout
SKILL.md:88

File Tree

2 files · 8.8 KB · 251 lines
Markdown 2f · 251L
├─ 📝 README.md Markdown 55L · 1.8 KB
└─ 📝 SKILL.md Markdown 196L · 7.0 KB

Security Positives

✓ No shell execution capabilities declared or inferred
✓ No file system access required
✓ No credential harvesting or environment variable enumeration
✓ No base64-encoded commands or obfuscation detected
✓ No hidden instructions in comments or documentation
✓ All network behavior is declared and central to the stated purpose
✓ Authentication is optional and supports anonymous mode
✓ Behavior is transparent: plan before, learn after, no side-channel operations