4 replies

bulldog20630405

giving agent exec perms

✅Solved

my agent seems to lost its exec perms when trying to do tasks
e..g

cat
grep
ls

cat
grep
ls

i have them already in the exec aproved file

this is the tools part of my openclaw.json

  "tools": {
    "profile": "messaging",
    "exec": {
      "host": "gateway",
      "security": "allowlist",
      "ask": "off"
    }
  },

  "tools": {
    "profile": "messaging",
    "exec": {
      "host": "gateway",
      "security": "allowlist",
      "ask": "off"
    }
  },

what profile do i need to set so the agent can work with running commands in the approve list?

Solution

tools.profile: "messaging"

tools.profile: "messaging"

is the reason: that profile does not include
exec
exec
(it only allows messaging + session tools). So the model never even gets the

exec

exec

tool, regardless of what’s in your exec-approvals file.

Fix options:

Option A (simplest): switch to

coding

coding

coding

coding

includes

group:runtime

group:runtime

(exec/process/bash) + fs/memory/sessions.

{
tools: {
profile: "coding",
exec: { host: "gateway", security: "allowlist", ask: "off" }
}
}

{
tools: {
profile: "coding",
exec: { host: "gateway", security: "allowlist", ask: "off" }
}
}

Option B: keep

messaging

messaging

, but explicitly allow runtime tools

Jump to solution

FotC�

Friends of the Crustacean 🦞🤝•10h ago•

4 replies

bulldog20630405

giving agent exec perms

✅Solved

my agent seems to lost its exec perms when trying to do tasks
e..g

cat
grep
ls

cat
grep
ls

i have them already in the exec aproved file

this is the tools part of my openclaw.json

  "tools": {
    "profile": "messaging",
    "exec": {
      "host": "gateway",
      "security": "allowlist",
      "ask": "off"
    }
  },

  "tools": {
    "profile": "messaging",
    "exec": {
      "host": "gateway",
      "security": "allowlist",
      "ask": "off"
    }
  },

what profile do i need to set so the agent can work with running commands in the approve list?

Solution

tools.profile: "messaging"

tools.profile: "messaging"

is the reason: that profile does not include
exec
exec
(it only allows messaging + session tools). So the model never even gets the

exec

exec

tool, regardless of what’s in your exec-approvals file.

Fix options:

Option A (simplest): switch to

coding

coding

coding

coding

includes

group:runtime

group:runtime

(exec/process/bash) + fs/memory/sessions.

{
tools: {
profile: "coding",
exec: { host: "gateway", security: "allowlist", ask: "off" }
}
}

{
tools: {
profile: "coding",
exec: { host: "gateway", security: "allowlist", ask: "off" }
}
}

Option B: keep

messaging

messaging

, but explicitly allow runtime tools

Jump to solution

giving agent exec perms

Option A (simplest): switch to
`coding`
`coding`

Option B: keep
`messaging`
`messaging`
, but explicitly allow runtime tools

Similar Threads

giving agent exec perms

Option A (simplest): switch to
`coding`
`coding`

Option B: keep
`messaging`
`messaging`
, but explicitly allow runtime tools

Similar Threads

Similar Threads

Similar Threads

giving agent exec perms

Option A (simplest): switch to codingcoding

Option B: keep messagingmessaging, but explicitly allow runtime tools

Similar Threads

giving agent exec perms

Option A (simplest): switch to codingcoding

Option B: keep messagingmessaging, but explicitly allow runtime tools

Similar Threads

Similar Threads

Similar Threads

Option A (simplest): switch to
`coding`
`coding`

Option B: keep
`messaging`
`messaging`
, but explicitly allow runtime tools

Option A (simplest): switch to
`coding`
`coding`

Option B: keep
`messaging`
`messaging`
, but explicitly allow runtime tools