suparank/publish

[Skill Scanner] Skill instructions include directives to hide actions from user All findings: [HIGH] autonomy_abuse: Skill instructions include directives to hide actions from user (BH009) [AITech 13.3] [HIGH] data_exfiltration: Outbound data post or form upload via curl/wget detected (NW002) [AITech 8.2.3] [HIGH] data_exfiltration: Outbound data post or form upload via curl/wget detected (NW002) [AITech 8.2.3] [HIGH] data_exfiltration: Outbound data post or form upload via curl/wget detected (NW002) [AITech 8.2.3] [HIGH] data_exfiltration: Outbound data post or form upload via curl/wget detected (NW002) [AITech 8.2.3] [HIGH] data_exfiltration: Outbound data post or form upload via curl/wget detected (NW002) [AITech 8.2.3] This skill's functionality is consistent with its stated purpose. It requires and reads sensitive credentials from local config files and uses them to call third-party APIs (WordPress, Ghost, fal.ai) and arbitrary webhooks — this is expected but high-sensitivity. The primary risks are: accidental credential leakage when building/executing shell/curl commands, reliance on an external image provider (fal.run), and sending data to arbitrary webhook URLs. There is no evidence of obfuscation, hardcoded secrets, or explicit malicious behavior (no download-execute chains or credential harvesting to unknown domains). Recommend careful implementation: ensure credentials are never printed or stored in logs, properly escape/validate payload contents, and restrict allowed webhook targets in high-security contexts. LLM verification: Functionally the skill is coherent with its stated purpose: it reads local credentials and article content, converts markdown to HTML/mobiledoc, posts to WordPress/Ghost, calls fal.ai to generate images, and sends webhooks. No hardcoded secrets, obfuscated payloads, or explicit backdoors are present. The primary security concerns are operational: use of inline shell commands that embed secrets (node -e with admin key, curl with Authorization headers) which can leak credentials via process listin