opencode-qa

Pass

Audited by Gen Agent Trust Hub on Jun 5, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local CLI tools including sqlite3, curl, jq, and tmux to perform automated testing and diagnostic tasks on the opencode application.
  • [COMMAND_EXECUTION]: The utility script scripts/lib/common.sh uses small inline Python or Bun snippets (python3 -c or bun -e) to dynamically identify available network ports for local server testing.
  • [DATA_EXFILTRATION]: The skill accesses the local application database at ~/.local/share/opencode/opencode.db to inspect session history. This access is performed read-only and is restricted to the local environment, with no evidence of data being transmitted to external or unauthorized domains.
  • [PROMPT_INJECTION]: An indirect prompt injection surface exists as the skill reads and displays message content from the local SQLite database. However, the risk is inherent to the skill's purpose as a debugger, and the scripts utilize single-quote escaping (oqa_sql_escape) to prevent SQL injection during local database queries.
  • [SAFE]: The skill intentionally isolates its execution environment by creating temporary XDG directories for data, config, and cache, ensuring that automated QA processes remain separated from production user data.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 5, 2026, 11:05 AM
Security Audit — agent-trust-hub — opencode-qa