Poll At Interval

User-Invocable: Yes (/poll <seconds> <tool> <condition>)

Controlled polling for async operations. Waits for a condition to be met, checking at specified intervals.

When to Use

•Waiting for Greptile review to complete
•Waiting for CI/CD status changes
•Any async operation that needs status polling
•Avoiding rate limit blocks from repeated tool calls

Syntax

code

/poll <interval_seconds> <tool_name> <condition_field> <expected_value>

Examples:

code

/poll 30 greptile:get_code_review status COMPLETED
/poll 60 gh:pr_checks conclusion success

How It Works

•Executes the specified tool
•Checks if condition is met
•If not met, waits interval_seconds
•Repeats until condition met or max iterations (default: 20)
•Returns final result

Parameters

Parameter	Description	Default
`interval`	Seconds between checks	Required
`tool`	MCP tool to call	Required
`condition_field`	JSON path to check	Required
`expected_value`	Value that signals completion	Required
`max_iterations`	Maximum poll attempts	20
`tool_params`	Parameters to pass to tool	{}

Bypass Hook Limits

This skill is whitelisted in combined-enforcement.py to avoid the 6-call limit. The hook recognizes polling patterns and allows controlled repetition.

Implementation

The skill spawns a background task that:

•Calls the tool with provided params
•Extracts condition_field from response
•Compares to expected_value
•If match: returns result
•If no match: sleeps interval, repeats

python

import time
import json

def poll_until(tool_name, tool_params, condition_field, expected_value,
               interval=30, max_iterations=20):
    """Poll a tool until condition is met."""
    for i in range(max_iterations):
        result = call_mcp_tool(tool_name, tool_params)

        # Extract nested field (e.g., "codeReview.status")
        value = result
        for key in condition_field.split('.'):
            value = value.get(key, {})

        if value == expected_value:
            return {"success": True, "iterations": i + 1, "result": result}

        if i < max_iterations - 1:
            time.sleep(interval)

    return {"success": False, "iterations": max_iterations, "last_result": result}

Integration with Hooks

To whitelist polling in combined-enforcement.py, add to check_mcp_script_requirement:

python

# Whitelist polling operations (controlled repetition)
POLLING_TOOLS = {
    "mcp__plugin_greptile_greptile__get_code_review",
    "mcp__plugin_greptile_greptile__list_code_reviews",
}

if tool_name in POLLING_TOOLS:
    # Check if this is a poll skill invocation
    if state.get("poll_skill_active"):
        return None  # Allow polling