opencode-delegate

Pass

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The scripts/relay.mjs file uses the Node.js child_process module (spawn, execFileSync, spawnSync) to execute the opencode CLI and git commands. This is necessary for the skill's purpose of delegating coding tasks and checking the status of the repository.\n- [PROMPT_INJECTION]: The skill is designed to ingest and process user-provided briefs and repository content to guide an external coding agent. This creates an indirect prompt injection surface.\n
  • Ingestion points: The relay script in scripts/relay.mjs reads task briefs from local files or standard input.\n
  • Boundary markers: The documentation (references/writing-the-brief.md) suggests using XML tags (e.g., <task>, <verification_loop>) to delineate instructions to the implementer.\n
  • Capability inventory: The relay.mjs script can execute shell commands and modify files via the OpenCode implementer.\n
  • Sanitization: The skill does not perform automated sanitization of the input briefs, relying instead on manual human review of the generated diffs as the primary security boundary.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 17, 2026, 07:09 PM
Security Audit — agent-trust-hub — opencode-delegate