execgo-agent-bridge
SkillAI & modelsRoute shell, runtime.command, runtime.script, mcp.call, cli.run, or task_graph.submit work through Secbot's ExecGo-backed control tools.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the execgo-agent-bridge skill
What this skill tells your AI
The instructions your AI receives, as published by iammm0/secbot in skills/custom/execgo-agent-bridge/SKILL.md and read by ahel’s review.
Use this skill when work should run through ExecGo instead of ad-hoc local execution. In this repository, the primary bridge is the execgo_action tool, and execute_command can also route through ExecGo when execgo=true or SECBOT_EXECGO_ENABLED=1 is active.
When To Use
- You need reliable task IDs, wait semantics, cancellation, timeout control, or runtime-backed execution.
- You want to submit
runtime.command,runtime.script,mcp.call,cli.run, ortask_graph.submitactions. - You need a structured smoke test before attempting a side-effecting ExecGo action.
Rules
- Prefer
execgo_actionover hand-written HTTP calls. - Use
mode=healthormode=toolsfirst if ExecGo availability is unknown. - Use a stable
action_idwhen subsequent waits or logs need correlation. - Start with
kind=os.noopfor connectivity checks before side-effecting actions. - For plain shell execution, prefer
execute_command; only force ExecGo when the job benefits from ExecGo task handling.
Tool Mapping In This Repo
execgo_action:mode=healthchecks ExecGo health.mode=toolslists adapter capabilities.- default
mode=actsubmits an action and waits by default.
execute_command:- pass
execgo=trueto route command execution through ExecGoruntime.command. - does not currently support
cwdorstdin_datain ExecGo mode.
- pass
Recommended Flow
- Call
execgo_actionwith{"mode":"health"}. - Call
execgo_actionwith{"mode":"tools"}if you need capability discovery. - Submit
os.noopfor a smoke test. - Submit the real action with a stable
action_idand structuredinput. - If the action is a simple shell command and fits the local bridge, use
execute_commandwithexecgo=trueinstead of building the JSON manually.
Examples
Smoke test via execgo_action:
{
"kind": "os.noop",
"action_id": "smoke-noop-1",
"input": {
"message": "hello execgo"
}
}
Runtime command via execgo_action:
{
"kind": "runtime.command",
"action_id": "runtime-echo-1",
"input": {
"program": "/bin/sh",
"args": ["-c", "echo hello execgo-runtime"],
"limits": {
"wall_time_ms": 30000
}
}
}
Shell command via execute_command:
{
"command": "uname -a",
"timeout": 30,
"execgo": true
}
Environment Notes
EXECGO_URLdefaults tohttp://127.0.0.1:8080.EXECGO_RUNTIME_URLdefaults tohttp://127.0.0.1:18080.EXECGO_EXECGOCLIcan point to a non-defaultexecgoclibinary.SECBOT_EXECGO_ENABLED=1makesexecute_commandprefer ExecGo by default.
References
references/action-contract.mdfor the action envelope and kinds.references/runtime-operations.mdfor runtime startup and readiness expectations.references/agent-specific-notes.mdfor notes about fitting the upstream skill into Secbot.references/troubleshooting.mdfor common failures.
Signals
- GitHub stars
- 74
- Forks
- 10
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
execgo-agent-bridge- Source
- github.com/iammm0/secbot