← Back to list

dsh-sandbox-arg-guard

Keeps a redundant sandbox-escalation argument from failing a tool call. Escalating tools (pwsh, bash, write, edit) advertise the full sandbox_permissions enum, but DSH only accepts a level strictly wider than the one already in effect — a check its own source calls 'deliberately not a schema constraint'. A model that reflexively passes the argument therefore picks the level it is already at and the call dies before running with 'sandbox escalation to "workspace-write" is not strictly wider than this call's current "workspace-write" mode', costing some models a whole turn of retries. This plugin registers one tools/execute waterfall listener and, only on that documented rejection and only when the arguments really carried an escalation field, re-issues the identical call once without it. Safe by DSH's own documentation: the rejection precedes any execution ('nothing has run'), and the corrected call cannot match again, so the retry is loop-free. Reproduced and verified end to end — before: isError true, the command never ran; after: the command's real output, isError false, one tool/call and one tool/result. Zero dependencies.

让「同级或更窄的 sandbox_permissions」不再让工具调用直接失败。会升级的工具（pwsh、bash、write、edit）都广告完整的 sandbox_permissions 枚举，但 DSH 只接受严格更宽于当前生效级别的请求——其源码自称这是「deliberately not a schema constraint」。于是反射式带上该参数的模型往往填它已经在的那个级别，调用在执行前就死掉："sandbox escalation to \"workspace-write\" is not strictly wider than this call's current \"workspace-write\" mode"，某些模型还会为此烧掉一整轮重试。本插件只注册一个 tools/execute waterfall 监听器，且仅在那一条文档化拒绝上、且参数里确实带了升级字段时，把同一个调用去掉该参数重投一次。安全性由 DSH 自己的文档保证：拒绝发生在任何执行之前（"nothing has run"），且改过的参数无法再次匹配，因此重投在结构上不成环。已端到端复现并验证——改造前 isError 为 true 且命令从未执行；改造后拿到命令的真实输出、isError 为 false，会话里只有一个 tool/call 与一个 tool/result。零依赖。