tailscale-policy-manager

Warn

Audited by Gen Agent Trust Hub on Mar 11, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides and describes the usage of multiple bash scripts (validate-policy.sh, apply-policy.sh, get-policy.sh) that use curl and python3 to interact with the Tailscale API. These scripts handle sensitive environment variables such as TS_API_KEY and TS_OAUTH_SECRET for network policy management.
  • [EXTERNAL_DOWNLOADS]: The skill utilizes external dependencies including the tailscale/gitops-acl-action GitHub Action and the hujsonfmt tool. These resources originate from Tailscale's official GitHub repository, a well-known technology service provider.
  • [DATA_EXFILTRATION]: The .claude/settings.local.json file grants WebFetch permissions for several non-whitelisted third-party domains including blog.gripdev.xyz, heywoodlh.io, and dev.l1qu1d.net. In a skill that handles critical network security configurations and API credentials, access to these unrelated personal domains presents a potential exfiltration vector.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) due to its configuration allowing data ingestion from untrusted sources via WebFetch to third-party domains. These ingestion points in .claude/settings.local.json lack boundary markers or sanitization, and the skill possesses significant capabilities (bash execution, API interaction) that could be exploited if malicious instructions are retrieved from these external sites.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 11, 2026, 07:16 PM