Adapted community discussion: Paraphrased and condensed from an anonymized public OpenClaw Discord support conversation from August 2026. Technical guidance was reviewed for OpenClaw 2026.9.3.
My agent can read and write files, but it cannot execute commands anymore. I must have changed something along the way. Where should I look to get execution back?
Reading, writing, and command execution are controlled separately. The usual causes are tool policy, sandbox policy, or an approval rule. Start with read-only checks:
openclaw --version
openclaw status --all
openclaw sandbox explain --json
openclaw approvals get
Also capture the exact execution error. Share only redacted output. Tokens, credentials, private paths, and host details should stay private.
The wording will narrow it down. If exec is missing, inspect allow and deny rules, because a deny wins. If the sandbox blocks it, inspect the sandbox tool policy. If an approval is pending or denied, inspect the exec approval rules. A command-not-found or filesystem permission error means execution is available and the selected host or environment is the problem.
Avoid switching everything to a broad full or elevated mode as a first step. That expands access and does not repair a policy denial.