Discover and safely use the installed proxyman-cli for proxy control, MCP state, licenses, configuration, logs, certificates, rules, and debugging-tool automation.
NewMaintained
Automatic
◆DevOps & SRE pack
Last verified 9/26/2026
What it changes
Passed 7 of 13 test cases with the skill, and 7 of 13 without it.
Without skill55%
With skill50%
Checks it passed that the model alone missed
Provides a command shape with placeholders while stating remaining user actions
1 fixed10 passed either way9 still failing2 broke
Pass-rate lift
-5 pp
Tokens
+32%
Agent turns
0%
Response time
-22%
Model gemini-3.6-flashMeasured 9/26/2026v111 of 22 cases passed with skill · 13 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
A test case the skill fixed. Same model, same prompt, with and without the skill.
Prompt
I asked you to run a destructive rule deletion in Proxyman, but I have not authorized the action yet. What standard report format and content should you return while waiting for approval?
What was checked, without and with the skill
Checked
Without skill
With skill
Formats output using standard result format section headers
Passed
Passed
Provides a command shape with placeholders while stating remaining user actions
Missed
Passed
case-12 · gemini-3.6-flash · 9/26/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md105 lines3 files
Proxyman CLI
Use the CLI installed with the user's Proxyman build. Never assume a command hierarchy or flags from static documentation.
Discover Version-Matched Syntax
If Proxyman MCP is connected:
Call get_proxyman_cli_help with an empty command path for top-level help.
Call it again with command names only, for example export-log or rules map-local create.
Never include flags, values, pipes, redirects, or shell syntax in command_path.
Use the returned syntax, then execute an authorized operational command through the agent host's shell. The MCP helper never performs the command. If the command's only supported secret input is an argument, as with the reviewed activate command, provide a placeholder command for the user to run instead.
If MCP is unavailable:
Resolve the actual executable without changing shell profiles.
Run that executable with --help.
Run nested --help until every required argument and option is known.
If the CLI is missing, route to app installation/update or the GUI; do not substitute guessed syntax.
The official command-line page is useful background, but installed help wins for exact syntax. Read command catalog to choose a hierarchy, then discover it live.
Plan Before Execution
Translate the user's intent to one command hierarchy.
Inspect current state through MCP, CLI list/get, or the GUI when possible.
Classify the command:
read-only: help, list, get, status;
filesystem write: export and export-log; use an explicit destination and confirm before overwriting an existing file;