Running commands behind the shell capability: the command policy, the process itself, and the truncation that keeps its output from swallowing the transcript.
Running commands behind the shell capability: the command policy, the process itself, and the truncation that keeps its output from swallowing the transcript.
Commands refused by default. This is a seatbelt, not a sandbox: it stops the catastrophic mistakes an agent actually makes, and a determined command can still work around any pattern. If you need a real boundary, use an allowlist or do not grant the shell.
Commands refused by default. This is a seatbelt, not a sandbox: it stops the catastrophic mistakes an agent actually makes, and a determined command can still work around any pattern. If you need a real boundary, use an allowlist or do not grant the shell.
(run-shell {:keys [shell-timeout-ms max-output strip-secrets?]
:or {strip-secrets? true}
:as opts}
{:keys [command dir]})Runs command through bash and returns its exit code plus combined
stdout/stderr. Never throws on a non-zero exit - a failing command is a
result the model should read, not an error.
Runs `command` through bash and returns its exit code plus combined stdout/stderr. Never throws on a non-zero exit - a failing command is a result the model should read, not an error.
(truncate s limit)Keeps the head and tail of s, since the interesting part of command output
is usually at one end or the other.
Keeps the head and tail of `s`, since the interesting part of command output is usually at one end or the other.
cljdoc builds & hosts documentation for Clojure/Script libraries
| Ctrl+k | Jump to recent docs |
| ← | Move to previous article |
| → | Move to next article |
| Ctrl+/ | Jump to the search field |