tiangong-wiki-skill

No explicit indicators of intentional malware (credential theft, backdoor installation, obfuscated payloads, or outbound exfiltration to unknown domains) are evident in this fragment. The code is primarily an ingestion/indexing and extraction pipeline. However, it processes potentially untrusted files (especially from a Synology download) and invokes system binaries (/usr/bin/mdls and /usr/bin/strings) on file paths, plus decompresses/parses ZIP-based Office documents with AdmZip without clear resource limits. The dominant security concern is elevated attack surface from untrusted file handling (path/symlink/path-traversal validation and archive resource exhaustion) rather than direct malicious behavior.

No direct evidence of embedded malware (no obvious backdoor, credential stealing, or network exfiltration endpoints) in this module. However, it materially increases operational/security impact through highly permissive Codex execution settings (default sandboxMode "danger-full-access", networkAccessEnabled=true, approvalPolicy="never"), injection of environment/credential material into the workflow runtime, and PATH modification that can influence binaries executed by spawned processes. The Windows spawn patch also reduces subprocess visibility. Overall: low likelihood of intentional malware in this specific fragment, but elevated security risk due to configuration and trust boundary assumptions (prompt and env can be attacker-influenced).

No clear malware behavior (no external C2/exfiltration, no reverse shell, no persistence) is present in this fragment. The dominant security concern is that it loads a caller-supplied JSON state file (via environment variable) and can serve its embedded credentials and file contents over HTTP on localhost after a weak session check. Additionally, the module increases supply-chain impact by executing embedded code via spawn(..., ['-e', script]). If used strictly for controlled tests with non-sensitive state, risk is limited; if used with sensitive state or exposed to other local processes, it can enable local data disclosure.

SUSPICIOUS: the stated wiki-management purpose is coherent, and the skill does not request disproportionate credentials or show explicit exfiltration. However, its core capability depends on an unverifiable external CLI with no install provenance, so the trust boundary is too weak to classify as benign.

No direct indicators of intentional malware are evident in this module (no obfuscation, no explicit credential/data theft, and no external exfiltration logic). However, it performs security-sensitive operations typical of dev tooling with elevated trust: it loads and applies local env files (including a parent-directory discovery mechanism and potentially out-of-scope absolute paths), spawns a child daemon process using that environment, and uses unvalidated host/port values from a local state file to perform HTTP health checks and to configure an /api reverse proxy. If an attacker can influence WIKI_ENV_FILE/WIKI_PATH or the local env/state files, the most plausible impact is SSRF-like internal reachability and unintended request forwarding via the dev proxy rather than classic malware behavior.