hivemind

SUSPICIOUS: the skill is purpose-aligned, but it is a thin wrapper around a third-party CLI/service that receives delegated task content and potentially code. The main risk is external trust and data exposure to opencode, not confirmed malware; installation provenance is partly official but muddied by package/repo transitions and an outdated package name in the instructions.

The code is a readable process wrapper and contains no clear malware indicators. It launches an external `opencode` service/command, performs a configurable health check, parses its output, and logs run metadata. The `--run` option is a concrete path traversal/arbitrary file-write issue because it is incorporated into a filesystem path without containment validation. The configurable URL and inherited environment are additional trust-boundary concerns. Restrict `--run` to a safe basename or verify the resolved path is inside `RUNS_DIR`, and consider restricting server URLs and child-process environment variables.

The code is a run-log summarizer with no evidence of malware, exfiltration, persistence, or command execution. It contains a moderate path traversal and local file disclosure risk because command-line identifiers are not constrained to safe filenames before constructing paths. Validate identifiers as basenames or resolve and verify that the resulting path remains inside RUNS_DIR. Treat log fields as untrusted if terminal output is security-sensitive.