Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Persona Sync

v1.1.3

在多个设备间通过 GitHub 私有仓库同步 AI 伙伴人格文件,确保记忆与个性不丢失。

0· 84·0 current·0 all-time
byHongyu Li@blueraincoatli
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Suspicious
medium confidence
!
Purpose & Capability
The skill claims to be an instruction-only persona sync helper, which fits the included sync.py, but the registry metadata lists no required binaries/envs while the code clearly depends on git and python3 (and optionally reads PERSONA_STORE_DIR). The missing declared dependencies is an incoherence a user should know about.
!
Instruction Scope
SKILL.md and SPEC.md describe local-only behavior and claim the token is never embedded in configs or logs. The script does read a local .gitauth, perform local git operations (init/fetch/pull/push), and write local files (memory.jsonl, state.json, sync.log). However the script also changes the user's global git credential helper (git config --global credential.helper store) and relies on git's credential store (~/.git-credentials), which is a cross-cutting system change outside the persona-store and is not obvious from the high-level guide.
Install Mechanism
No remote install or downloaded code is executed; this is an instruction/script bundle (no install spec). The code is local and uses subprocess calls to git — low installation risk, but runtime side-effects still matter.
!
Credentials
The skill requests no env vars in registry metadata, yet the script supports PERSONA_STORE_DIR and requires git to function and expects a GitHub PAT in a local .gitauth file. It will cause the PAT to be stored by git's credential helper in ~/.git-credentials (global), which the SKILL.md asserts will be mode 0600 but the script does not enforce or set file permissions for that file. Asking users to place a PAT in a local file and then changing global git behavior is disproportionate to what the metadata advertised.
!
Persistence & Privilege
The skill does not run automatically and always:false is correct, but it modifies global git configuration (credential.helper = store) which persists beyond the persona-store and affects all git operations on the host. This persistent side-effect is not clearly called out in the metadata and increases blast radius for credential exposure.
What to consider before installing
This skill largely does what it says — it syncs persona files to a GitHub private repo using a small Python script — but there are important surprises you should consider before installing: - Dependency mismatch: The registry metadata claims no required binaries, but the script requires git and python3. Ensure those are present and trusted. - Global git config change: The script runs `git config --global credential.helper store`, which permanently changes your global Git behavior (it makes Git store credentials in ~/.git-credentials). That affects all repositories on the machine. - Plaintext credential storage: Using 'store' writes credentials to ~/.git-credentials (plaintext). The SKILL.md/SPEC.md claim 0600 permissions and 'secure storage', but the script does not set or verify permissions on ~/.git-credentials. Consider whether you accept storing a PAT in plaintext on this machine. - Alternatives to consider: use a repo-specific approach (SSH deploy keys or a local-only credential helper), use a PAT with minimal scope, enable 2FA, or run the sync inside a restricted environment (container or dedicated user account) to limit exposure. - What to check if you want to proceed: review scripts/sync.py yourself, confirm you are comfortable with git --global changes, manually run init to see what gets written, and after first run verify permissions of ~/.git-credentials and whether git config --global was changed. Because of the global config/credential storage side effects and the metadata inconsistencies, treat this skill as suspicious until you verify those behaviors and accept the risks.

Like a lobster shell, security has layers — review code before you run it.

latestvk97egxx5s2f85fb39fc74348xh83q53b

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Comments