[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M79DKSCSXHZT8XVBMS4JCF6M":3,"skill-same:01M79DKSCSXHZT8XVBMS4JCF6M":129},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":27,"commitSha":28,"skillDirectory":29,"forksCount":30,"repositoryHtmlUrl":31,"repositoryHomepageUrl":32,"ownerHtmlUrl":33,"ownerBlogUrl":34,"ownerDescription":35,"updatedAt":36,"install":37,"resources":50},"01M79DKSCSXHZT8XVBMS4JCF6M","agent-platform-alert-configuration",null,"Configures best-practice alerting policies for Google Cloud Vertex AI \u002F Agent Platform agents on Agent Runtime. Use when analyzing, writing, or deploying alerting policies to monitor agent latency, error rates, and quality metrics (response quality, tool use, hallucination). Also use when provisioning online monitors for quality evaluation, or analyzing live metrics traffic footprints. NOTE: This skill currently only works for the Agent Runtime. Don't use for configuring general GCP alert policies or non-agent GCP alerting policies.","hamzabellouch\u002Fagent-skills","hamzabellouch","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F237856042?v=4","01M79BRYW7XAMK3YDWQEMTWGWS","Hamza Bellouch","01M79BRYWXQV3JK2TGVTX8GN2A",8,"MIT","https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: agent-platform-alert-configuration\nmetadata:\n  category: AiAndMachineLearning\ndescription: >-\n  Configures best-practice alerting policies for Google Cloud Vertex AI \u002F Agent\n  Platform agents on Agent Runtime. Use when analyzing, writing, or deploying\n  alerting policies to monitor agent latency, error rates, and quality metrics\n  (response quality, tool use, hallucination). Also use when provisioning online\n  monitors for quality evaluation, or analyzing live metrics traffic footprints.\n  NOTE: This skill currently only works for the Agent Runtime. Don't use for\n  configuring general GCP alert policies or non-agent GCP alerting policies.\nallowed-tools: terraform gcloud python\n---\n\n# Agent Platform Alert Configuration\n\nThis skill provides dynamic threshold alerting configurations for Google Cloud \u002F\nVertex AI Reasoning Engines (Agent Platform container deployments) using\nextended 1-week lookback retention baselines. Standard static thresholds (e.g.,\n\"latency > 2s\") cause excessive alert noise for AI agents. Dynamic PromQL\nbaselines solve this.\n\n## Safety & Confirmation Tiers (CRITICAL)\n\nBefore executing any commands or writing configurations on behalf of the user,\nyou MUST adhere to the following safety tiers based on the action requested:\n\n1.  **Tier R: Read-only (`check_telemetry.py`)**\n    *   **Rule**: No confirmation needed. You may execute these scripts\n        immediately to inspect the telemetry status of the Reasoning Engine.\n2.  **Tier B: Billing & Resource Creation (`create_online_monitor.py` \u002F\n    provisioning)**\n    *   **Rule**: **Explicit User Confirmation Required**. These actions incur\n        additional billing charges and create cloud resources. The agent MUST\n        ask the user directly for approval before proceeding.\n\n--------------------------------------------------------------------------------\n\n## CRITICAL RULES\n\n*   **Always configure both Reliability and Quality alerting policies** for the\n    target agent (6 policies in total):\n    *   **For Reliability Monitoring**: You MUST configure exactly three\n        alerting policies:\n        1.  **Latency** (anomaly monitoring)\n        2.  **Error Rate - Fast Burn SLO** (1-Hour Window)\n        3.  **Error Rate - Slow Burn SLO** (3-Day Window)\n    *   **For Quality Monitoring**: You MUST configure exactly three alerting\n        policies:\n        1.  **Final Response Quality**\n        2.  **Tool Use Quality**\n        3.  **Hallucination**\n*   **Online Monitor Provisioning & Cost Warning**: Quality alerting policies\n    rely on metrics exported by Online Monitors. You MUST ensure the Online\n    Monitor is provisioned for the agent and telemetry is enabled:\n    -   [ ] **Ask for Approval**: Both Online Monitors and Telemetry incur\n        separate billing charges. Before provisioning them, you MUST warn the\n        user about these extra costs. If not pre-approved in the prompt, you\n        MUST ask a direct question in your response requesting\n        confirmation\u002Fapproval to proceed (e.g., \"Please confirm if you approve\n        the extra billing costs for the Online Monitor and Telemetry to\n        proceed.\").\n    -   [ ] **Verify Telemetry First**: Before generating any alerting policy\n        plan or provisioning Online Monitors, you MUST always verify the\n        telemetry status of the Reasoning Engine first using the\n        `check_telemetry.py` script as detailed in\n        [Verify Telemetry Status](#verify-telemetry-status) below.\n    -   [ ] **Follow the Guide**: Follow the step-by-step instructions in the\n        [Online Monitor & Telemetry Provisioning](#online-monitor--telemetry-provisioning)\n        section below.\n*   **Brand New Agents (No Traffic History)**: When setting up alerts for a\n    brand new agent, you MUST explicitly ask the user what traffic pattern they\n    expect (Steady, Seasonal, or Bursty) in your response. If immediate setup is\n    requested, ask the question but proceed using the default Steady\u002FConsistent\n    (Short-Window Z-Score) pattern. Follow\n    [no_historical_traffic_data.md](references\u002Fno_historical_traffic_data.md).\n*   **PromQL for Reliability (No MQL or Threshold Filters)**: For the 3\n    reliability metrics, you MUST use `condition_prometheus_query_language` with\n    PromQL. Do **NOT** use MQL or standard `condition_threshold`.\n*   **Standard Threshold Filters for Agent Quality**: For the 3 agent quality\n    metrics, you MUST use standard `condition_threshold` filters matching the\n    monitored resource type `aiplatform.googleapis.com\u002FOnlineEvaluator` and\n    metric type `aiplatform.googleapis.com\u002Fonline_evaluator\u002Fscores`. Do **NOT**\n    use PromQL.\n*   **Install Terraform if Necessary**: You should use terraform to deploy and\n    must install terraform if you can't find a valid install.\n*   **Terraform Only**: Write the generated observability configuration ONLY as\n    Terraform (`.tf`) files (e.g., `alerts.tf`, `variables.tf`).\n*   **Dynamic Multi-Resource Alerting (No Single-Resource Pinning)**: You MUST\n    NOT hardcode specific agent IDs or resource name filters (e.g.,\n    `{reasoning_engine_id=\"[AGENT_ID]\"}` or\n    `metric.labels.agent_resource_name=\"[AGENT_NAME]\"`) in alerting conditions\n    unless explicitly requested. Alerting policies must be written to cover all\n    active agents in the project dynamically:\n    *   **For Reliability Metrics using PromQL**: ALWAYS use grouping\n        aggregations (`by (reasoning_engine_id)`) instead of filtering to a\n        single ID. This allows a single alert policy to dynamically track each\n        reasoning engine instance separately.\n    *   **For Quality Metrics using Standard Threshold Filters**: Omit the\n        `agent_resource_name` filter entirely. Configure the condition filter to\n        only target the monitored resource type\n        (`aiplatform.googleapis.com\u002FOnlineEvaluator`) and metric type\n        (`aiplatform.googleapis.com\u002Fonline_evaluator\u002Fscores`) globally for the\n        project.\n*   **Check for Pre-existing Policies**: Avoid creating duplicate alert policies\n    for a reasoning engine: scan the target directory or workspace to see if a\n    policy already exists that targets the same metrics using aggregations\n    grouped by `reasoning_engine_id`.\n*   **Metric Scope Discovery & Project Inference**: Centralize alert policies in\n    a Metric Scope (scoping project) to save costs. Identify if a scope is used\n    and where policies should live by checking:\n    1.  **GCP CLI Check**: Run `gcloud beta monitoring metrics-scopes list\n        projects\u002F[PROJECT_ID]`. If a parent scope\n        `locations\u002Fglobal\u002FmetricsScopes\u002F[SCOPING_PROJECT_ID]` is returned, a\n        Metric Scope is active; deploy policies there.\n    2.  **Infrastructure as Code Scan**: Search Terraform configurations for\n        `google_monitoring_monitored_project` resources and extract the scoping\n        project from the `metrics_scope` attribute.\n    3.  **Ambiguity Fallback**: If unable to determine, ask the user: \"Are you\n        using a multi-project Cloud Monitoring Metric Scope? If so, what is the\n        scoping project ID?\" Deploy policies to the deduced scoping project\n        (setting the `project` attribute in HCL), or default to the local\n        project.\n*   **Directory Inference**: Deploy configuration files to target Terraform or\n    SRE folders (e.g. `monitoring\u002F`, `ops\u002F`, `sre\u002F`). Use tools to locate where\n    alert policies or state pointers exist in the project, rather than blindly\n    writing to the current working directory.\n*   **Notification Channels**: By default, never configure any notification\n    channels without user input. If the user explicitly provides a notification\n    channel in their prompt, configure the alerts to use it. If no notification\n    channel is provided, you MUST explicitly ask the user in your final response\n    if they would like to configure notification channels. **This is a mandatory\n    question and you MUST NOT omit it from your response.** **IMPORTANT** Do NOT\n    make assumptions about notification channels. If you search the codebase for\n    a notification channel you must ALWAYS confirm with the user before using\n    it.\n*   **Plain English Response**: You MUST include a plain English explanation for\n    what the alerts do in your response. This must explain in plain English what\n    the alert measures, how the algorithm works, and what a trigger indicates.\n*   **Avoid Recursive Directory Operations**: You MUST NOT run recursive listing\n    or search commands (such as `ls -R`, `find .`, or raw recursive `grep`) from\n    the google3 workspace root, as this will hang your session. Always target\n    specific subdirectories.\n*   **Background Task Cleanup**: You MUST check the status of all background\n    tasks that you spawn. Before completing your execution and returning your\n    final response, you MUST terminate or kill any active or hanging background\n    tasks (using the `manage_task` tool with action `kill`).\n\n--------------------------------------------------------------------------------\n\n## Algorithm Selection & Policy Mapping Process\n\nAlerting policies for reasoning engine agents MUST map to the correct algorithms\nto ensure statistical stability and prevent alert noise or blind spots based on\ndata classes:\n\n*   **Latency**: Follows workload traffic pattern (Steady -> Z-Score; Seasonal\n    -> Seasonal Decomposition; Bursty -> Moving Averages).\n*   **Error Rate**: ALWAYS use **Multi-Window Multi-Burn Rate SLOs** (or\n    ratio-based static thresholds). Error rate is naturally sparse (normally\n    `0`). When standard deviation is `0`, Z-score computation is mathematically\n    unstable (division-by-zero or NaN), causing false alert storms.\n\nTo resolve the workload traffic pattern (Seasonal, Steady, or Bursty), follow\nthe instructions corresponding to the availability of historical metrics data:\n\n*   **Case 1: No historical metrics data available (e.g., brand new agent)**:\n    You MUST read and follow:\n    [no_historical_traffic_data.md](references\u002Fno_historical_traffic_data.md)\n*   **Case 2: Historical metrics data available (e.g., active agent with\n    traffic)**: You MUST read and follow:\n    [has_historical_traffic_data.md](references\u002Fhas_historical_traffic_data.md)\n\n--------------------------------------------------------------------------------\n\n## Telemetry Metrics and PromQL Examples\n\nAll raw telemetry metrics for the Agent Platform are cumulative **counters**.\nBecause we monitor their rates or quantiles, we can optimize the PromQL queries\nby using longer range windows (e.g., `[1w]`) for historical averages instead of\nexpensive `avg_over_time` subqueries.\n\nSignal         | Raw Metric                                  | Type    | Description\n:------------- | :------------------------------------------ | :------ | :----------\n**Latency**    | `reasoning_engine_request_latencies_bucket` | Counter | Histogram bucket of request latencies\n**Error Rate** | `reasoning_engine_request_count`            | Counter | Cumulative count of requests\n\n--------------------------------------------------------------------------------\n\nFor the specific PromQL queries corresponding to each algorithm, you MUST read\nand follow: [promql_queries.md](references\u002Fpromql_queries.md)\n\n--------------------------------------------------------------------------------\n\n## Agent Quality Metrics (Online Monitor)\n\nAll agent quality evaluation metrics are exported by Online Monitors to the\nmonitored resource type `aiplatform.googleapis.com\u002FOnlineEvaluator` under the\nmetric type `aiplatform.googleapis.com\u002Fonline_evaluator\u002Fscores`.\n\n### Metric Details & Aligners\n\nBecause the scores metric is of value type `DISTRIBUTION`, standard mean-based\nPromQL or arithmetic `ALIGN_MEAN` aligners are unsupported. You MUST use a\npercentile aligner (typically `ALIGN_PERCENTILE_50` to evaluate the median\nscore) within the `aggregations` block of your `condition_threshold`.\n\nSignal                           | Metric Name (`evaluation_metric_name`) | Target Threshold    | Recommended Aligner\n:------------------------------- | :------------------------------------- | :------------------ | :------------------\n**Final Response Quality**       | `final_response_quality_v1`            | `\u003C 0.8` (or custom) | `ALIGN_PERCENTILE_50`\n**Tool Use Quality**             | `tool_use_quality_v1`                  | `\u003C 0.8` (or custom) | `ALIGN_PERCENTILE_50`\n**Hallucination (Groundedness)** | `hallucination_v1`                     | `\u003C 0.9` (or custom) | `ALIGN_PERCENTILE_50`\n\n### Metric Filter Example\n\nWhen configuring a quality alert policy in Terraform, use the following filter\nexpression structure:\n\n```filter\nresource.type=\"aiplatform.googleapis.com\u002FOnlineEvaluator\"\nAND metric.type=\"aiplatform.googleapis.com\u002Fonline_evaluator\u002Fscores\"\nAND metric.labels.evaluation_metric_name=\"[METRIC_NAME]\"\n```\n\n### Online Monitor & Telemetry Provisioning\n\nQuality metrics are generated by the Online Monitor by evaluating trace data\nexported to Cloud Trace. If telemetry is disabled on the reasoning engine, no\ntraces are sent, and the quality metrics will remain empty.\n\n#### Prerequisites & Dependencies\n\nBefore executing any scripts in this skill (such as `check_telemetry.py` or\n`create_online_monitor.py`), you MUST install the required dependencies in your\nenvironment. Run this command first:\n\n```bash\npip install -r scripts\u002Frequirements.txt\n```\n\n#### Verify Telemetry Status\n\nBefore generating any alerting policies, proposing a plan, or provisioning\nOnline Monitors, you MUST always check if the agent is ready to export traces by\nrunning the telemetry checking script:\n\n*   **Mandatory Command**: `python3 scripts\u002Fcheck_telemetry.py --project-id\n    \"[PROJECT_ID]\" --agent-resource-name \"[AGENT_RESOURCE_NAME]\"`\n    *   **Note on Parameters**: The `[AGENT_RESOURCE_NAME]` parameter MUST be\n        the full resource path format\n        `projects\u002F\u003Cproject_id>\u002Flocations\u002F\u003Clocation>\u002FreasoningEngines\u002F\u003Cagent_id>`\n        (e.g. `projects\u002Fgcp-prod\u002Flocations\u002Fus-central1\u002FreasoningEngines\u002F556677`)\n        and not just the agent ID itself.\n    *   **Dependency Failures**: If package imports or dependency installation\n        fails, try your best to resolve the issues (e.g., by verifying package\n        installation) and run the script again with the specific project-id and\n        agent-resource-name parameters. If you cannot run the script\n        successfully due to missing dependencies, you MUST still attempt to run\n        it first and then include the complete, pre-populated execution command\n        in your final response as a verification plan (this applies even if the\n        user has pre-approved provisioning).\n*   **Detailed Enablement Instructions**: For details on the required\n    environment variables, Terraform setups, and project dependencies, you MUST\n    read and follow:\n    [telemetry_enablement.md](references\u002Ftelemetry_enablement.md).\n\n#### Provision the Online Monitor\n\nBecause Online Monitors cannot be configured via Terraform, run the provisioning\nscript to generate the required metrics:\n\n*   **Command**: `python3 scripts\u002Fcreate_online_monitor.py --project-id\n    \"[PROJECT_ID]\" --agent-resource-name \"[AGENT_RESOURCE_NAME]\"\n    [--sampling-percentage [PERCENTAGE]]`\n*   **Sampling Rate Recommendation**: For production agents, configure a\n    conservative sampling percentage (default: **10%**) to control LLM\n    evaluation costs. For details, refer to\n    [Continuous evaluation with online monitors](https:\u002F\u002Fdocs.cloud.google.com\u002Fgemini-enterprise-agent-platform\u002Foptimize\u002Fevaluation\u002Fevaluate-online).\n\n#### Formatting the Execution Plan\n\nIf execution fails (e.g. due to sandbox restrictions or permissions), you MUST\nplan it as a concrete next step. To do this, you MUST explicitly include a\nsection titled **\"Online Monitor Provisioning Plan\"** or **\"Execution Plan\"** in\nyour final response, containing the exact concrete python execution command with\nall parameter values (such as project ID, region, and agent resource name) fully\npopulated. Do not merely state that the user should run it.\n\nYou MUST format the plan exactly as follows:\n\n### Execution Plan: Online Monitor Provisioning\n\n**Online Monitor Provisioning Command:**\n\n```bash\npython3 scripts\u002Fcreate_online_monitor.py \\\n  --project-id \"[PROJECT_ID]\" \\\n  --agent-resource-name \"projects\u002F[PROJECT_ID]\u002Flocations\u002F[LOCATION]\u002FreasoningEngines\u002F[AGENT_ID]\" \\\n  --sampling-percentage [PERCENTAGE]\n```\n\n**Verify Telemetry Command (Optional fallback):**\n\n```bash\npython3 scripts\u002Fcheck_telemetry.py \\\n  --project-id \"[PROJECT_ID]\" \\\n  --agent-resource-name \"projects\u002F[PROJECT_ID]\u002Flocations\u002F[LOCATION]\u002FreasoningEngines\u002F[AGENT_ID]\"\n```\n\n--------------------------------------------------------------------------------\n\n## Tooling Scripts\n\nUse the following scripts to resolve duplicates and validate configs before\npresenting or applying Terraform changes:\n\n1.  **Duplicate Check & Merge**: Checks for pre-existing alerts in the target\n    folder to ensure changes are merged in-place rather than appended:\n    *   Command: `python3 scripts\u002Fvalidate_config.py --directory [TARGET_TF_DIR]\n        --engine-var \"${var.reasoning_engine_id}\"`\n2.  **Config Linting**: Validates PromQL grammar, matching engine labels, and\n    HCL structure:\n    *   Command: `python3 scripts\u002Fvalidate_config.py --file [PATH_TO_TF_FILE]`\n    *   **Self-Correction Loop**: If validation fails (exits non-zero or outputs\n        errors), you MUST read the command output, locate the line\u002Ffile\n        containing the lint error, analyze the PromQL syntax or Terraform HCL\n        issue, apply adjustments in-place, and re-run the `validate_config.py\n        --file` validation. Repeat this loop until the validation script passes\n        successfully.\n\n--------------------------------------------------------------------------------\n\n## Gotchas & Behavioral Corrections\n\n*   **Duration Buffers (Transient Glitches)**: To avoid alerts firing on\n    transient spikes, use duration\u002Fretest window buffers appropriately:\n    *   **Reliability Metrics (PromQL \u002F Cloud Monitoring)**:\n        *   For short-lookback alerts querying data under 25 hours (e.g.,\n            Short-Window Z-Score, Moving Averages, Fast Burn SLO), ALWAYS use a\n            `duration = \"300s\"` (5 minutes) buffer to filter out transient cold\n            start\u002Fdeployment spikes.\n        *   For long-lookback alerts querying data longer than 25 hours (e.g.,\n            Long-Window Z-Score, Seasonal Decomposition, Slow Burn SLO),\n            duration\u002Fretest windows are disabled by the platform. You must **not\n            set a duration** (omit it entirely).\n    *   **Quality Metrics (Standard Filters \u002F Online Monitor)**:\n        *   Always use a `duration = \"300s\"` (5 minutes) buffer to filter out\n            transient scoring dips or evaluation outliers caused by temporary\n            LLM judge congestion, or edge-case query outliers.\n*   **Dynamic Baseline Adaptation Blind Spot**: Explain to users that dynamic\n    statistical Z-score thresholds compare current rates to a moving statistical\n    baseline. If a system degrades slowly over days, the standard baseline curve\n    adapts to this slow drift, making standard Z-score alerts blind to\n    persistent slow errors. Recommend a hard static threshold alert in parallel\n    for strict SLA enforcement.\n*   **Seasonal Decomposition Double Alerting**: The agent MUST ONLY configure\n    seasonal decomposition alert policies to track spikes (e.g., latency spikes)\n    OR drops AND MUST NOT use dual-direction checks (like absolute deviation).\n    Explain this limitation to the user: comparing to a historical offset (e.g.,\n    `offset 1w`) the alert policy triggers twice if tracking both directions\n    (once for the anomaly, and once 1 week later when the anomaly becomes the\n    baseline). To prevent this, the generated policy MUST only track either\n    spikes (using `>`) or drops (using `\u003C`), avoiding using `abs()`.\n*   **Raw Error Boundaries**: Explain that raw error counts or absolute failed\n    request count boundaries do not scale under changing traffic throughput.\n    Recommend ratio-based error rate alerts instead.\n*   **Safe Threshold Modulation E2E Validation**: When verifying a dynamic\n    metric threshold policy end-to-end, do NOT attempt to force real platform\n    errors. Instead, deploy the alert policy with standard safe bounds (Z-score\n    multiplier > 15), then temporarily update standard deviation Z-score limits\n    to a negative value (e.g. > -3) to trigger\u002Fverify the \"Firing\" state before\n    reverting. Always get confirmation before taking this action proactively.\n*   **Expected Script Failures**:\n    *   `validate_config.py --directory` exiting with code 1: Parse the JSON\n        output for duplicate resource targets. Perform in-place upgrade edits,\n        then re-check until it passes with 0.\n    *   **Script Execution Failures & Self-Correction**: If the execution of\n        utility scripts (such as `check_telemetry.py`,\n        `create_online_monitor.py`, or `analyze_traffic.py`) fails unexpectedly,\n        you MUST read and inspect the stdout\u002Fstderr logs or error output.\n        Analyze the error message (e.g., connection timeouts, invalid\n        permissions, or missing resources) and attempt to dynamically correct\n        parameters (such as verifying or correcting the region, project ID, or\n        resource name format) and retry execution before escalating or falling\n        back to manual plans.\n*   **Distribution Metric Aligner Constraint**: Standard `ALIGN_MEAN` cannot be\n    applied to `DELTA` distribution metrics like `online_evaluator\u002Fscores`. You\n    MUST use percentile-based aligners (like `ALIGN_PERCENTILE_50`) to reduce\n    the score distribution into a comparable numeric stream.\n\n--------------------------------------------------------------------------------\n\n## Supporting Links\n\n*   [Continuous evaluation with online monitors](https:\u002F\u002Fdocs.cloud.google.com\u002Fgemini-enterprise-agent-platform\u002Foptimize\u002Fevaluation\u002Fevaluate-online)\n*   [Agent Platform Quality Metrics](https:\u002F\u002Fdocs.cloud.google.com\u002Fgemini-enterprise-agent-platform\u002Fmodels\u002Frubric-metric-details)\n*   [Google Cloud Alerting Policies Guide](https:\u002F\u002Fcloud.google.com\u002Fmonitoring\u002Falerts)\n*   [Google Cloud Monitoring PromQL Documentation](https:\u002F\u002Fcloud.google.com\u002Fmonitoring\u002Fpromql)\n","# Agent Platform Alert Configuration\n\nThis skill provides dynamic threshold alerting configurations for Google Cloud \u002F\nVertex AI Reasoning Engines (Agent Platform container deployments) using\nextended 1-week lookback retention baselines. Standard static thresholds (e.g.,\n\"latency > 2s\") cause excessive alert noise for AI agents. Dynamic PromQL\nbaselines solve this.\n\n## Safety & Confirmation Tiers (CRITICAL)\n\nBefore executing any commands or writing configurations on behalf of the user,\nyou MUST adhere to the following safety tiers based on the action requested:\n\n1.  **Tier R: Read-only (`check_telemetry.py`)**\n    *   **Rule**: No confirmation needed. You may execute these scripts\n        immediately to inspect the telemetry status of the Reasoning Engine.\n2.  **Tier B: Billing & Resource Creation (`create_online_monitor.py` \u002F\n    provisioning)**\n    *   **Rule**: **Explicit User Confirmation Required**. These actions incur\n        additional billing charges and create cloud resources. The agent MUST\n        ask the user directly for approval before proceeding.\n\n--------------------------------------------------------------------------------\n\n## CRITICAL RULES\n\n*   **Always configure both Reliability and Quality alerting policies** for the\n    target agent (6 policies in total):\n    *   **For Reliability Monitoring**: You MUST configure exactly three\n        alerting policies:\n        1.  **Latency** (anomaly monitoring)\n        2.  **Error Rate - Fast Burn SLO** (1-Hour Window)\n        3.  **Error Rate - Slow Burn SLO** (3-Day Window)\n    *   **For Quality Monitoring**: You MUST configure exactly three alerting\n        policies:\n        1.  **Final Response Quality**\n        2.  **Tool Use Quality**\n        3.  **Hallucination**\n*   **Online Monitor Provisioning & Cost Warning**: Quality alerting policies\n    rely on metrics exported by Online Monitors. You MUST ensure the Online\n    Monitor is provisioned for the agent and telemetry is enabled:\n    -   [ ] **Ask for Approval**: Both Online Monitors and Telemetry incur\n        separate billing charges. Before provisioning them, you MUST warn the\n        user about these extra costs. If not pre-approved in the prompt, you\n        MUST ask a direct question in your response requesting\n        confirmation\u002Fapproval to proceed (e.g., \"Please confirm if you approve\n        the extra billing costs for the Online Monitor and Telemetry to\n        proceed.\").\n    -   [ ] **Verify Telemetry First**: Before generating any alerting policy\n        plan or provisioning Online Monitors, you MUST always verify the\n        telemetry status of the Reasoning Engine first using the\n        `check_telemetry.py` script as detailed in\n        [Verify Telemetry Status](#verify-telemetry-status) below.\n    -   [ ] **Follow the Guide**: Follow the step-by-step instructions in the\n        [Online Monitor & Telemetry Provisioning](#online-monitor--telemetry-provisioning)\n        section below.\n*   **Brand New Agents (No Traffic History)**: When setting up alerts for a\n    brand new agent, you MUST explicitly ask the user what traffic pattern they\n    expect (Steady, Seasonal, or Bursty) in your response. If immediate setup is\n    requested, ask the question but proceed using the default Steady\u002FConsistent\n    (Short-Window Z-Score) pattern. Follow\n    [no_historical_traffic_data.md](references\u002Fno_historical_traffic_data.md).\n*   **PromQL for Reliability (No MQL or Threshold Filters)**: For the 3\n    reliability metrics, you MUST use `condition_prometheus_query_language` with\n    PromQL. Do **NOT** use MQL or standard `condition_threshold`.\n*   **Standard Threshold Filters for Agent Quality**: For the 3 agent quality\n    metrics, you MUST use standard `condition_threshold` filters matching the\n    monitored resource type `aiplatform.googleapis.com\u002FOnlineEvaluator` and\n    metric type `aiplatform.googleapis.com\u002Fonline_evaluator\u002Fscores`. Do **NOT**\n    use PromQL.\n*   **Install Terraform if Necessary**: You should use terraform to deploy and\n    must install terraform if you can't find a valid install.\n*   **Terraform Only**: Write the generated observability configuration ONLY as\n    Terraform (`.tf`) files (e.g., `alerts.tf`, `variables.tf`).\n*   **Dynamic Multi-Resource Alerting (No Single-Resource Pinning)**: You MUST\n    NOT hardcode specific agent IDs or resource name filters (e.g.,\n    `{reasoning_engine_id=\"[AGENT_ID]\"}` or\n    `metric.labels.agent_resource_name=\"[AGENT_NAME]\"`) in alerting conditions\n    unless explicitly requested. Alerting policies must be written to cover all\n    active agents in the project dynamically:\n    *   **For Reliability Metrics using PromQL**: ALWAYS use grouping\n        aggregations (`by (reasoning_engine_id)`) instead of filtering to a\n        single ID. This allows a single alert policy to dynamically track each\n        reasoning engine instance separately.\n    *   **For Quality Metrics using Standard Threshold Filters**: Omit the\n        `agent_resource_name` filter entirely. Configure the condition filter to\n        only target the monitored resource type\n        (`aiplatform.googleapis.com\u002FOnlineEvaluator`) and metric type\n        (`aiplatform.googleapis.com\u002Fonline_evaluator\u002Fscores`) globally for the\n        project.\n*   **Check for Pre-existing Policies**: Avoid creating duplicate alert policies\n    for a reasoning engine: scan the target directory or workspace to see if a\n    policy already exists that targets the same metrics using aggregations\n    grouped by `reasoning_engine_id`.\n*   **Metric Scope Discovery & Project Inference**: Centralize alert policies in\n    a Metric Scope (scoping project) to save costs. Identify if a scope is used\n    and where policies should live by checking:\n    1.  **GCP CLI Check**: Run `gcloud beta monitoring metrics-scopes list\n        projects\u002F[PROJECT_ID]`. If a parent scope\n        `locations\u002Fglobal\u002FmetricsScopes\u002F[SCOPING_PROJECT_ID]` is returned, a\n        Metric Scope is active; deploy policies there.\n    2.  **Infrastructure as Code Scan**: Search Terraform configurations for\n        `google_monitoring_monitored_project` resources and extract the scoping\n        project from the `metrics_scope` attribute.\n    3.  **Ambiguity Fallback**: If unable to determine, ask the user: \"Are you\n        using a multi-project Cloud Monitoring Metric Scope? If so, what is the\n        scoping project ID?\" Deploy policies to the deduced scoping project\n        (setting the `project` attribute in HCL), or default to the local\n        project.\n*   **Directory Inference**: Deploy configuration files to target Terraform or\n    SRE folders (e.g. `monitoring\u002F`, `ops\u002F`, `sre\u002F`). Use tools to locate where\n    alert policies or state pointers exist in the project, rather than blindly\n    writing to the current working directory.\n*   **Notification Channels**: By default, never configure any notification\n    channels without user input. If the user explicitly provides a notification\n    channel in their prompt, configure the alerts to use it. If no notification\n    channel is provided, you MUST explicitly ask the user in your final response\n    if they would like to configure notification channels. **This is a mandatory\n    question and you MUST NOT omit it from your response.** **IMPORTANT** Do NOT\n    make assumptions about notification channels. If you search the codebase for\n    a notification channel you must ALWAYS confirm with the user before using\n    it.\n*   **Plain English Response**: You MUST include a plain English explanation for\n    what the alerts do in your response. This must explain in plain English what\n    the alert measures, how the algorithm works, and what a trigger indicates.\n*   **Avoid Recursive Directory Operations**: You MUST NOT run recursive listing\n    or search commands (such as `ls -R`, `find .`, or raw recursive `grep`) from\n    the google3 workspace root, as this will hang your session. Always target\n    specific subdirectories.\n*   **Background Task Cleanup**: You MUST check the status of all background\n    tasks that you spawn. Before completing your execution and returning your\n    final response, you MUST terminate or kill any active or hanging background\n    tasks (using the `manage_task` tool with action `kill`).\n\n--------------------------------------------------------------------------------\n\n## Algorithm Selection & Policy Mapping Process\n\nAlerting policies for reasoning engine agents MUST map to the correct algorithms\nto ensure statistical stability and prevent alert noise or blind spots based on\ndata classes:\n\n*   **Latency**: Follows workload traffic pattern (Steady -> Z-Score; Seasonal\n    -> Seasonal Decomposition; Bursty -> Moving Averages).\n*   **Error Rate**: ALWAYS use **Multi-Window Multi-Burn Rate SLOs** (or\n    ratio-based static thresholds). Error rate is naturally sparse (normally\n    `0`). When standard deviation is `0`, Z-score computation is mathematically\n    unstable (division-by-zero or NaN), causing false alert storms.\n\nTo resolve the workload traffic pattern (Seasonal, Steady, or Bursty), follow\nthe instructions corresponding to the availability of historical metrics data:\n\n*   **Case 1: No historical metrics data available (e.g., brand new agent)**:\n    You MUST read and follow:\n    [no_historical_traffic_data.md](references\u002Fno_historical_traffic_data.md)\n*   **Case 2: Historical metrics data available (e.g., active agent with\n    traffic)**: You MUST read and follow:\n    [has_historical_traffic_data.md](references\u002Fhas_historical_traffic_data.md)\n\n--------------------------------------------------------------------------------\n\n## Telemetry Metrics and PromQL Examples\n\nAll raw telemetry metrics for the Agent Platform are cumulative **counters**.\nBecause we monitor their rates or quantiles, we can optimize the PromQL queries\nby using longer range windows (e.g., `[1w]`) for historical averages instead of\nexpensive `avg_over_time` subqueries.\n\nSignal         | Raw Metric                                  | Type    | Description\n:------------- | :------------------------------------------ | :------ | :----------\n**Latency**    | `reasoning_engine_request_latencies_bucket` | Counter | Histogram bucket of request latencies\n**Error Rate** | `reasoning_engine_request_count`            | Counter | Cumulative count of requests\n\n--------------------------------------------------------------------------------\n\nFor the specific PromQL queries corresponding to each algorithm, you MUST read\nand follow: [promql_queries.md](references\u002Fpromql_queries.md)\n\n--------------------------------------------------------------------------------\n\n## Agent Quality Metrics (Online Monitor)\n\nAll agent quality evaluation metrics are exported by Online Monitors to the\nmonitored resource type `aiplatform.googleapis.com\u002FOnlineEvaluator` under the\nmetric type `aiplatform.googleapis.com\u002Fonline_evaluator\u002Fscores`.\n\n### Metric Details & Aligners\n\nBecause the scores metric is of value type `DISTRIBUTION`, standard mean-based\nPromQL or arithmetic `ALIGN_MEAN` aligners are unsupported. You MUST use a\npercentile aligner (typically `ALIGN_PERCENTILE_50` to evaluate the median\nscore) within the `aggregations` block of your `condition_threshold`.\n\nSignal                           | Metric Name (`evaluation_metric_name`) | Target Threshold    | Recommended Aligner\n:------------------------------- | :------------------------------------- | :------------------ | :------------------\n**Final Response Quality**       | `final_response_quality_v1`            | `\u003C 0.8` (or custom) | `ALIGN_PERCENTILE_50`\n**Tool Use Quality**             | `tool_use_quality_v1`                  | `\u003C 0.8` (or custom) | `ALIGN_PERCENTILE_50`\n**Hallucination (Groundedness)** | `hallucination_v1`                     | `\u003C 0.9` (or custom) | `ALIGN_PERCENTILE_50`\n\n### Metric Filter Example\n\nWhen configuring a quality alert policy in Terraform, use the following filter\nexpression structure:\n\n```filter\nresource.type=\"aiplatform.googleapis.com\u002FOnlineEvaluator\"\nAND metric.type=\"aiplatform.googleapis.com\u002Fonline_evaluator\u002Fscores\"\nAND metric.labels.evaluation_metric_name=\"[METRIC_NAME]\"\n```\n\n### Online Monitor & Telemetry Provisioning\n\nQuality metrics are generated by the Online Monitor by evaluating trace data\nexported to Cloud Trace. If telemetry is disabled on the reasoning engine, no\ntraces are sent, and the quality metrics will remain empty.\n\n#### Prerequisites & Dependencies\n\nBefore executing any scripts in this skill (such as `check_telemetry.py` or\n`create_online_monitor.py`), you MUST install the required dependencies in your\nenvironment. Run this command first:\n\n```bash\npip install -r scripts\u002Frequirements.txt\n```\n\n#### Verify Telemetry Status\n\nBefore generating any alerting policies, proposing a plan, or provisioning\nOnline Monitors, you MUST always check if the agent is ready to export traces by\nrunning the telemetry checking script:\n\n*   **Mandatory Command**: `python3 scripts\u002Fcheck_telemetry.py --project-id\n    \"[PROJECT_ID]\" --agent-resource-name \"[AGENT_RESOURCE_NAME]\"`\n    *   **Note on Parameters**: The `[AGENT_RESOURCE_NAME]` parameter MUST be\n        the full resource path format\n        `projects\u002F\u003Cproject_id>\u002Flocations\u002F\u003Clocation>\u002FreasoningEngines\u002F\u003Cagent_id>`\n        (e.g. `projects\u002Fgcp-prod\u002Flocations\u002Fus-central1\u002FreasoningEngines\u002F556677`)\n        and not just the agent ID itself.\n    *   **Dependency Failures**: If package imports or dependency installation\n        fails, try your best to resolve the issues (e.g., by verifying package\n        installation) and run the script again with the specific project-id and\n        agent-resource-name parameters. If you cannot run the script\n        successfully due to missing dependencies, you MUST still attempt to run\n        it first and then include the complete, pre-populated execution command\n        in your final response as a verification plan (this applies even if the\n        user has pre-approved provisioning).\n*   **Detailed Enablement Instructions**: For details on the required\n    environment variables, Terraform setups, and project dependencies, you MUST\n    read and follow:\n    [telemetry_enablement.md](references\u002Ftelemetry_enablement.md).\n\n#### Provision the Online Monitor\n\nBecause Online Monitors cannot be configured via Terraform, run the provisioning\nscript to generate the required metrics:\n\n*   **Command**: `python3 scripts\u002Fcreate_online_monitor.py --project-id\n    \"[PROJECT_ID]\" --agent-resource-name \"[AGENT_RESOURCE_NAME]\"\n    [--sampling-percentage [PERCENTAGE]]`\n*   **Sampling Rate Recommendation**: For production agents, configure a\n    conservative sampling percentage (default: **10%**) to control LLM\n    evaluation costs. For details, refer to\n    [Continuous evaluation with online monitors](https:\u002F\u002Fdocs.cloud.google.com\u002Fgemini-enterprise-agent-platform\u002Foptimize\u002Fevaluation\u002Fevaluate-online).\n\n#### Formatting the Execution Plan\n\nIf execution fails (e.g. due to sandbox restrictions or permissions), you MUST\nplan it as a concrete next step. To do this, you MUST explicitly include a\nsection titled **\"Online Monitor Provisioning Plan\"** or **\"Execution Plan\"** in\nyour final response, containing the exact concrete python execution command with\nall parameter values (such as project ID, region, and agent resource name) fully\npopulated. Do not merely state that the user should run it.\n\nYou MUST format the plan exactly as follows:\n\n### Execution Plan: Online Monitor Provisioning\n\n**Online Monitor Provisioning Command:**\n\n```bash\npython3 scripts\u002Fcreate_online_monitor.py \\\n  --project-id \"[PROJECT_ID]\" \\\n  --agent-resource-name \"projects\u002F[PROJECT_ID]\u002Flocations\u002F[LOCATION]\u002FreasoningEngines\u002F[AGENT_ID]\" \\\n  --sampling-percentage [PERCENTAGE]\n```\n\n**Verify Telemetry Command (Optional fallback):**\n\n```bash\npython3 scripts\u002Fcheck_telemetry.py \\\n  --project-id \"[PROJECT_ID]\" \\\n  --agent-resource-name \"projects\u002F[PROJECT_ID]\u002Flocations\u002F[LOCATION]\u002FreasoningEngines\u002F[AGENT_ID]\"\n```\n\n--------------------------------------------------------------------------------\n\n## Tooling Scripts\n\nUse the following scripts to resolve duplicates and validate configs before\npresenting or applying Terraform changes:\n\n1.  **Duplicate Check & Merge**: Checks for pre-existing alerts in the target\n    folder to ensure changes are merged in-place rather than appended:\n    *   Command: `python3 scripts\u002Fvalidate_config.py --directory [TARGET_TF_DIR]\n        --engine-var \"${var.reasoning_engine_id}\"`\n2.  **Config Linting**: Validates PromQL grammar, matching engine labels, and\n    HCL structure:\n    *   Command: `python3 scripts\u002Fvalidate_config.py --file [PATH_TO_TF_FILE]`\n    *   **Self-Correction Loop**: If validation fails (exits non-zero or outputs\n        errors), you MUST read the command output, locate the line\u002Ffile\n        containing the lint error, analyze the PromQL syntax or Terraform HCL\n        issue, apply adjustments in-place, and re-run the `validate_config.py\n        --file` validation. Repeat this loop until the validation script passes\n        successfully.\n\n--------------------------------------------------------------------------------\n\n## Gotchas & Behavioral Corrections\n\n*   **Duration Buffers (Transient Glitches)**: To avoid alerts firing on\n    transient spikes, use duration\u002Fretest window buffers appropriately:\n    *   **Reliability Metrics (PromQL \u002F Cloud Monitoring)**:\n        *   For short-lookback alerts querying data under 25 hours (e.g.,\n            Short-Window Z-Score, Moving Averages, Fast Burn SLO), ALWAYS use a\n            `duration = \"300s\"` (5 minutes) buffer to filter out transient cold\n            start\u002Fdeployment spikes.\n        *   For long-lookback alerts querying data longer than 25 hours (e.g.,\n            Long-Window Z-Score, Seasonal Decomposition, Slow Burn SLO),\n            duration\u002Fretest windows are disabled by the platform. You must **not\n            set a duration** (omit it entirely).\n    *   **Quality Metrics (Standard Filters \u002F Online Monitor)**:\n        *   Always use a `duration = \"300s\"` (5 minutes) buffer to filter out\n            transient scoring dips or evaluation outliers caused by temporary\n            LLM judge congestion, or edge-case query outliers.\n*   **Dynamic Baseline Adaptation Blind Spot**: Explain to users that dynamic\n    statistical Z-score thresholds compare current rates to a moving statistical\n    baseline. If a system degrades slowly over days, the standard baseline curve\n    adapts to this slow drift, making standard Z-score alerts blind to\n    persistent slow errors. Recommend a hard static threshold alert in parallel\n    for strict SLA enforcement.\n*   **Seasonal Decomposition Double Alerting**: The agent MUST ONLY configure\n    seasonal decomposition alert policies to track spikes (e.g., latency spikes)\n    OR drops AND MUST NOT use dual-direction checks (like absolute deviation).\n    Explain this limitation to the user: comparing to a historical offset (e.g.,\n    `offset 1w`) the alert policy triggers twice if tracking both directions\n    (once for the anomaly, and once 1 week later when the anomaly becomes the\n    baseline). To prevent this, the generated policy MUST only track either\n    spikes (using `>`) or drops (using `\u003C`), avoiding using `abs()`.\n*   **Raw Error Boundaries**: Explain that raw error counts or absolute failed\n    request count boundaries do not scale under changing traffic throughput.\n    Recommend ratio-based error rate alerts instead.\n*   **Safe Threshold Modulation E2E Validation**: When verifying a dynamic\n    metric threshold policy end-to-end, do NOT attempt to force real platform\n    errors. Instead, deploy the alert policy with standard safe bounds (Z-score\n    multiplier > 15), then temporarily update standard deviation Z-score limits\n    to a negative value (e.g. > -3) to trigger\u002Fverify the \"Firing\" state before\n    reverting. Always get confirmation before taking this action proactively.\n*   **Expected Script Failures**:\n    *   `validate_config.py --directory` exiting with code 1: Parse the JSON\n        output for duplicate resource targets. Perform in-place upgrade edits,\n        then re-check until it passes with 0.\n    *   **Script Execution Failures & Self-Correction**: If the execution of\n        utility scripts (such as `check_telemetry.py`,\n        `create_online_monitor.py`, or `analyze_traffic.py`) fails unexpectedly,\n        you MUST read and inspect the stdout\u002Fstderr logs or error output.\n        Analyze the error message (e.g., connection timeouts, invalid\n        permissions, or missing resources) and attempt to dynamically correct\n        parameters (such as verifying or correcting the region, project ID, or\n        resource name format) and retry execution before escalating or falling\n        back to manual plans.\n*   **Distribution Metric Aligner Constraint**: Standard `ALIGN_MEAN` cannot be\n    applied to `DELTA` distribution metrics like `online_evaluator\u002Fscores`. You\n    MUST use percentile-based aligners (like `ALIGN_PERCENTILE_50`) to reduce\n    the score distribution into a comparable numeric stream.\n\n--------------------------------------------------------------------------------\n\n## Supporting Links\n\n*   [Continuous evaluation with online monitors](https:\u002F\u002Fdocs.cloud.google.com\u002Fgemini-enterprise-agent-platform\u002Foptimize\u002Fevaluation\u002Fevaluate-online)\n*   [Agent Platform Quality Metrics](https:\u002F\u002Fdocs.cloud.google.com\u002Fgemini-enterprise-agent-platform\u002Fmodels\u002Frubric-metric-details)\n*   [Google Cloud Alerting Policies Guide](https:\u002F\u002Fcloud.google.com\u002Fmonitoring\u002Falerts)\n*   [Google Cloud Monitoring PromQL Documentation](https:\u002F\u002Fcloud.google.com\u002Fmonitoring\u002Fpromql)\n",{"name":5,"metadata":24,"description":7,"allowed-tools":26},{"category":25},"AiAndMachineLearning","terraform gcloud python","main","e905af21e5f1e1c3ca605e9656c3a7930360a84e","Programming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration",1,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills","https:\u002F\u002Fhamzabellouch.github.io\u002F","https:\u002F\u002Fgithub.com\u002Fhamzabellouch","hamzabellouch.github.io","CEO & Founder of @tknoglobal | Researcher in Technology & Applied Sciences.","2026-10-17 03:20:19",{"source":8,"skillName":5,"commands":38,"prompt":48,"zipAvailable":49},[39,42,45],{"tool":40,"command":41},"npx","npx skills add hamzabellouch\u002Fagent-skills --skill agent-platform-alert-configuration",{"tool":43,"command":44},"bunx","bunx skills add hamzabellouch\u002Fagent-skills --skill agent-platform-alert-configuration",{"tool":46,"command":47},"pnpm","pnpm dlx skills add hamzabellouch\u002Fagent-skills --skill agent-platform-alert-configuration","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002FSKILL.md",true,[51,57,61,65,69,73,79,83,87,91,97,101,105,109,113,117,121,125],{"path":52,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":55,"sourceHtmlUrl":56},"assets\u002Falerts_initial_duplicate.tf","ASSET","资产文件",655,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fassets\u002Falerts_initial_duplicate.tf",{"path":58,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":59,"sourceHtmlUrl":60},"assets\u002Fdraft_invalid_query.tf",411,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fassets\u002Fdraft_invalid_query.tf",{"path":62,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":63,"sourceHtmlUrl":64},"assets\u002Fmock_bursty.json",32486,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fassets\u002Fmock_bursty.json",{"path":66,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":67,"sourceHtmlUrl":68},"assets\u002Fmock_seasonal.json",34332,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fassets\u002Fmock_seasonal.json",{"path":70,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":71,"sourceHtmlUrl":72},"assets\u002Fmock_steady.json",35829,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fassets\u002Fmock_steady.json",{"path":74,"resourceType":75,"resourceTypeDesc":76,"sizeBytes":77,"sourceHtmlUrl":78},"references\u002Fhas_historical_traffic_data.md","REFERENCE","参考文档",3824,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Freferences\u002Fhas_historical_traffic_data.md",{"path":80,"resourceType":75,"resourceTypeDesc":76,"sizeBytes":81,"sourceHtmlUrl":82},"references\u002Fno_historical_traffic_data.md",3875,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Freferences\u002Fno_historical_traffic_data.md",{"path":84,"resourceType":75,"resourceTypeDesc":76,"sizeBytes":85,"sourceHtmlUrl":86},"references\u002Fpromql_queries.md",5041,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Freferences\u002Fpromql_queries.md",{"path":88,"resourceType":75,"resourceTypeDesc":76,"sizeBytes":89,"sourceHtmlUrl":90},"references\u002Ftelemetry_enablement.md",2513,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Freferences\u002Ftelemetry_enablement.md",{"path":92,"resourceType":93,"resourceTypeDesc":94,"sizeBytes":95,"sourceHtmlUrl":96},"scripts\u002Fanalyze_traffic_test.py","SCRIPT","脚本",6459,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fscripts\u002Fanalyze_traffic_test.py",{"path":98,"resourceType":93,"resourceTypeDesc":94,"sizeBytes":99,"sourceHtmlUrl":100},"scripts\u002Fanalyze_traffic.py",10753,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fscripts\u002Fanalyze_traffic.py",{"path":102,"resourceType":93,"resourceTypeDesc":94,"sizeBytes":103,"sourceHtmlUrl":104},"scripts\u002Fcheck_telemetry_test.py",3917,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fscripts\u002Fcheck_telemetry_test.py",{"path":106,"resourceType":93,"resourceTypeDesc":94,"sizeBytes":107,"sourceHtmlUrl":108},"scripts\u002Fcheck_telemetry.py",3265,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fscripts\u002Fcheck_telemetry.py",{"path":110,"resourceType":93,"resourceTypeDesc":94,"sizeBytes":111,"sourceHtmlUrl":112},"scripts\u002Fcreate_online_monitor_test.py",6116,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fscripts\u002Fcreate_online_monitor_test.py",{"path":114,"resourceType":93,"resourceTypeDesc":94,"sizeBytes":115,"sourceHtmlUrl":116},"scripts\u002Fcreate_online_monitor.py",5234,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fscripts\u002Fcreate_online_monitor.py",{"path":118,"resourceType":93,"resourceTypeDesc":94,"sizeBytes":119,"sourceHtmlUrl":120},"scripts\u002Frequirements.txt",60,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fscripts\u002Frequirements.txt",{"path":122,"resourceType":93,"resourceTypeDesc":94,"sizeBytes":123,"sourceHtmlUrl":124},"scripts\u002Fvalidate_config_test.py",11430,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fscripts\u002Fvalidate_config_test.py",{"path":126,"resourceType":93,"resourceTypeDesc":94,"sizeBytes":127,"sourceHtmlUrl":128},"scripts\u002Fvalidate_config.py",10860,"https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FProgramming\u002FAI_Engineering_and_Platforms\u002FAgents_and_LLM_Platforms\u002Fagent-platform-alert-configuration\u002Fscripts\u002Fvalidate_config.py",[130,137,144,151,158,165],{"skillNo":131,"name":132,"nameZh":6,"description":133,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":134,"categories":135,"keywords":136,"publishedAt":19},"01M79FHH2H815A9TAP6WNKH4Q8","security-audit","Comprehensive security guidance, source code audit, vulnerability assessment, and penetration test review for codebases, APIs, microservices, CLI tools, libraries, and LLM-backed applications. Triggers on requests to audit a codebase, conduct security reviews, find vulnerabilities or flaws, pen-test source code, or when the user asks in Arabic (فحص أمني، تدقيق الكود أمنياً، مراجعة الثغرات، فحص أمان المشروع). Supports Guidance mode (focused review) and Full Audit mode (deterministic 6-phase harness).","https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FSecurity\u002FSecurity_Audit_and_Code_Review\u002Fsecurity-audit\u002FSKILL.md",[],[],{"skillNo":138,"name":139,"nameZh":6,"description":140,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":141,"categories":142,"keywords":143,"publishedAt":19},"01M79FHBGXVBY6VQ1X7ETAZMF1","malware-analysis-reversing","Production malware analysis workflows, reverse engineering methodologies, isolated sandbox triage, static\u002Fdynamic binary analysis, memory forensics, and YARA detection rule authoring.","https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FSecurity\u002FOffensive_Security_and_Pentesting\u002Fmalware-analysis-reversing\u002FSKILL.md",[],[],{"skillNo":145,"name":146,"nameZh":6,"description":147,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":148,"categories":149,"keywords":150,"publishedAt":19},"01M79FH6EACDR801TZCE4R8SAC","application-security-pentesting","Production Application Security (AppSec) standards, OWASP Top 10 vulnerability remediation, SAST\u002FDAST pipeline integration, secure code review, threat modeling, and defensive security posture.","https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FSecurity\u002FOffensive_Security_and_Pentesting\u002Fapplication-security-pentesting\u002FSKILL.md",[],[],{"skillNo":152,"name":153,"nameZh":6,"description":154,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":155,"categories":156,"keywords":157,"publishedAt":19},"01M79FH0VKVT44FK923K520ZQA","oauth2-oidc-security-flows","Implement enterprise OAuth 2.1 and OpenID Connect (OIDC) authorization code flow with Proof Key for Code Exchange (PKCE), state and nonce verification, refresh token rotation, token revocation, and centralized IdP federation (Keycloak, Auth0, Okta). Trigger when building secure authentication, single sign-on (SSO), or third-party delegated authorization.","https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FSecurity\u002FIdentity_and_Access_Management_IAM\u002Foauth2-oidc-security-flows\u002FSKILL.md",[],[],{"skillNo":159,"name":160,"nameZh":6,"description":161,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":162,"categories":163,"keywords":164,"publishedAt":19},"01M79FGVVB3R4T88XXD8FRSV5B","jwt-session-management-hardening","Harden JWT-based authentication and session lifecycles using asymmetric signing (RS256\u002FEdDSA), JWKS rotation, strict claims validation, Redis-backed sliding window sessions, revocation lists, and secure HttpOnly cookie storage. Trigger when implementing token validation, session revocation, or hardening API authentication.","https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FSecurity\u002FIdentity_and_Access_Management_IAM\u002Fjwt-session-management-hardening\u002FSKILL.md",[],[],{"skillNo":166,"name":167,"nameZh":6,"description":168,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":169,"categories":170,"keywords":171,"publishedAt":19},"01M79FGNCPZ85CVEY0B3NE4H3B","soc2-compliance-automation","Best practices for implementing SOC 2 Type II compliance controls, continuous evidence collection, infrastructure policy-as-code, and automated security monitoring. Use when automating Trust Services Criteria (Security, Availability, Confidentiality, Processing Integrity, Privacy), audit logging, IAM least privilege, or CI\u002FCD security gates.","https:\u002F\u002Fgithub.com\u002Fhamzabellouch\u002Fagent-skills\u002Fblob\u002Fmain\u002FSecurity\u002FGovernance_Compliance_and_Legal_Tech\u002Fsoc2-compliance-automation\u002FSKILL.md",[],[]]