Is dboy233/amap safe?

https://github.com/openclaw/skills/tree/main/skills/dboy233/amap

93
SAFE

The dboy233/amap skill is a legitimate, well-structured integration for the Amap (高德地图) geolocation API, providing weather, POI search, routing, and geocoding services for Chinese geographic data. No malicious code, prompt injection, hidden instructions, or unauthorized file access was detected — the installation was entirely clean with only the two expected files added to disk. The sole noteworthy concern is inherent to the skill's purpose: all location queries are processed by Alibaba/AutoNavi servers in China, raising data sovereignty considerations for users outside China or those with location-privacy requirements.

Category Scores

Prompt Injection 95/100 · 30%
Data Exfiltration 83/100 · 25%
Code Execution 98/100 · 20%
Clone Behavior 90/100 · 10%
Canary Integrity 100/100 · 10%
Behavioral Reasoning 88/100 · 5%

Findings (5)

LOW Location Data Processed by Alibaba/AutoNavi Infrastructure -12

Every geographic query made through this skill — including user addresses, coordinates, travel routes, and city names — is transmitted to restapi.amap.com, operated by AutoNavi (a wholly-owned Alibaba subsidiary). This is expected and inherent behavior for a maps API skill, but users should be aware their location data is processed on Chinese servers subject to PRC data regulations and government access obligations. This is a data sovereignty concern, not malicious exfiltration.

LOW API Key Transmitted in Plaintext URL Query Parameters -5

The AMAP_KEY is interpolated directly into curl URL strings as a query parameter (key=$AMAP_KEY). This is standard practice for the Amap Web Service API but means the key appears in HTTP access logs, reverse proxy logs, tcpdump captures, and shell history. An agent with shell logging could inadvertently leak the key.

INFO Clean Installation — No Side Effects 0

The installation process was entirely clean. Only the two expected files were written to disk. All network traffic was to GitHub for the sparse checkout. No unexpected processes were spawned, no persistent connections established, and no files were modified outside the skill directory.

INFO Canary File Reads Attributed to Audit Framework 0

The monitoring logs record reads of canary files (.env, .ssh/id_rsa, .aws/credentials, .npmrc, .docker/config.json, gcloud credentials) at two points in the session. Timing analysis confirms both sets of reads are from the oathe audit framework: the first batch (audit 1771941167.221) occurs 5 seconds before the git clone begins, and the second batch (audit 1771941184.691) aligns with the post-scan phase. The skill itself contains no code capable of reading these files.

INFO Chinese-Language Skill — Content Verified Clean 0

SKILL.md is written in Simplified Chinese. Full content was reviewed and translated. The skill contains only legitimate API documentation, appropriate trigger conditions for geographic queries, and standard curl examples. No hidden instructions in any language were found.