project-governance

SUSPICIOUS. The skill is internally coherent as a large project-governance orchestrator and does not show clear malicious exfiltration or credential-harvesting behavior, but its scope is very broad and it delegates trust to several external/custom tools and companion skills whose provenance is only partly verifiable here. Main risk is operational power and transitive trust, not confirmed malware.

This module is a generic manifest-driven local command runner. It does not show overt malicious behavior (no credential theft, no network activity, no obfuscation, no persistence), but it has a high-impact security property: it will execute whatever executable/argv is declared in a resolved JSON manifest via subprocess, with only a simple mutability/--authorized gate and conditional validation of parameter values. The dominant supply-chain risk is integrity/trust of resolve_task() and the manifest contents (operation['command'], parameters, and mutability). Without manifest authenticity controls, a compromised/attacker-influenced manifest can lead to arbitrary command execution on the host.

The code is not obviously malicious by itself: it is a transparent release workflow orchestrator with strong input validation and no built-in backdoors, network exfiltration, or obfuscated payloads. The main security risk is architectural: it executes arbitrary, repository-controlled hook commands (gates/freeze/deploy/verify/migration/hotfix) via subprocess.run with the full environment, so compromise or tampering of the release-workflow.json/config refs can lead to arbitrary code execution and data theft/exfiltration by those hooks. Tar extraction into a temp dir and broad env propagation further widen the impact of any malicious hook/controller content. Overall: no clear malware indicators in this module, but high privilege execution makes supply-chain/hook integrity critical.

This module is an orchestrator for repository governance auditing: it inventories and hashes Markdown files, then executes external Node/uv tooling to validate governance and scan lifecycle targets, producing a structured JSON report. There is no direct evidence of credential theft, persistence, or in-module data exfiltration. The primary security concern is execution/supply-chain integrity: it may execute an mdq.py script discovered from user home directories without integrity pinning, and it runs external commands whose behavior is outside this module. If the environment (home directories, invoked scripts, or their dependencies) is tampered with, the orchestration could be leveraged to run attacker-controlled code.