A critique says OpenCode’s agent design breaks KV cache and weakens security
JFPuget · x · 2026-08-04
A long critique says OpenCode’s design breaks key security assumptions
A linked essay argues that OpenCode, a popular open-source coding agent with 161k GitHub stars, should be avoided because of serious system and security flaws.
- The author says OpenCode is effectively a llm | bash web-stack tool and that its failures are structural, not incidental.
- The post focuses on two buckets of problems: annoying UX issues and more alarming security issues.
- The critique says OpenCode breaks KV cache behavior and has a weak security posture around command execution, permissions, and redirection handling.
- An update note says some behavior has changed in newer commits, including disabled tool-call pruning by default, removal of cross-session persistence for command-prefix permissions, and path validation for bash redirections.
- The writer still claims many concerns remain and says a prior GitHub stale-bot closure of a security report is especially troubling.
More from coding & agent
- New demo videos were assembled entirely autonomously, with no human visibility until the end — jasonkneen · 2026-08-04
- OpenHands adds ToolShield to software-agent-sdk, cutting attack success to 7–10% — shi_weiyan · 2026-08-04
- Split AI work across multiple harnesses, not one all-purpose session — EXM7777 · 2026-08-04
- Every shows how voice plus AI agents can patch bugs and write work — every · 2026-08-04
- Applied AI’s real edge is workflow logic, not raw engineering skill — brandon_galang · 2026-08-04
- Browser MCP scanner flags shell exec, secrets, and unsafe deserialization — KookyTax5493 · 2026-08-04