[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7A8J4XVFTQ82W1RAMFK5CTB":3,"skill-same:01M7A8J4XVFTQ82W1RAMFK5CTB":144},{"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":44,"commitSha":45,"skillDirectory":46,"forksCount":47,"repositoryHtmlUrl":48,"repositoryHomepageUrl":6,"ownerHtmlUrl":49,"ownerBlogUrl":6,"ownerDescription":50,"updatedAt":51,"install":52,"resources":65},"01M7A8J4XVFTQ82W1RAMFK5CTB","fis-experiment-deployer",null,"Provisions AWS Fault Injection Service (FIS) experiment templates with production-safe defaults: action targets (tags, ARNs, filters), fault actions across EC2 (stop-instances, send-api-error, terminate), ECS (stop-task), network (blackhole\u002Flatency\u002Floss via SSM), Aurora (failover-db-cluster), Lambda (invoke-async), CloudWatch alarm-based stop conditions, IAM execution role with FIS trust + least-privilege tag-scoped permissions, experiment logging to S3 + CloudWatch Logs, and latest primitives (Aurora failover, EKS pod disruption, network faults). Emits READY_TO_DEPLOY \u002F PREREQUISITES_MISSING with verified targets, actions, stop conditions, and copy-pasteable fis + cloudwatch commands. Use when provisioning chaos experiments for game days, DR drills, resilience validation, or pre-prod gates. Triggers: FIS, fault injection, chaos engineering, experiment template, stop condition, Aurora failover, EKS pod kill, network blackhole, game day.","jhkchan\u002Faws-agent-skills","jhkchan","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1083902?v=4","01M7A2SVQYP9GQTPTDXP5JKYQ4","Jacky Chan","01M7A2SVS117YWCTZABB75HZR1",4,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: fis-experiment-deployer\ndescription: 'Provisions AWS Fault Injection Service (FIS) experiment templates with production-safe defaults: action targets (tags, ARNs, filters), fault actions across EC2 (stop-instances, send-api-error, terminate), ECS (stop-task), network (blackhole\u002Flatency\u002Floss via SSM), Aurora (failover-db-cluster), Lambda (invoke-async), CloudWatch alarm-based stop conditions, IAM execution role with FIS trust + least-privilege tag-scoped permissions, experiment logging to S3 + CloudWatch Logs, and latest primitives (Aurora failover, EKS pod disruption, network faults). Emits READY_TO_DEPLOY \u002F PREREQUISITES_MISSING with verified targets, actions, stop conditions, and copy-pasteable fis + cloudwatch commands. Use when provisioning chaos experiments for game days, DR drills, resilience validation, or pre-prod gates. Triggers: FIS, fault injection, chaos engineering, experiment template, stop condition, Aurora failover, EKS pod kill, network blackhole, game day.'\nlicense: Apache-2.0\ncompatibility: Agent runtime that reads SKILL.md (Claude Code, Cursor, Windsurf, Codex, Gemini). Live provisioning uses AWS CLI v2 with fis (create-experiment-template, get-experiment-template, start-experiment), iam (create-role, put-role-policy), cloudwatch (describe-alarms), s3 (create-bucket, put-bucket-policy) and logs (create-log-group). Works with Terraform aws_fis_experiment_template \u002F aws_iam_role and CloudFormation AWS::FIS::ExperimentTemplate.\nmetadata:\n  domain: aws-cloudops\n  complexity: high\n  requires_llm: 'true'\n  phase: '1'\n  supports_pipeline: 'true'\n  entry_point: 'false'\n  family: Management\n  task_type: deploy\n  skill_class: capability\n  lifecycle_status: active\n  verdict_shape: READY_TO_DEPLOY | PREREQUISITES_MISSING\n  when_to_use: Provisioning an FIS experiment template for a game day, DR drill, or resilience gate; configuring CloudWatch alarms as stop conditions; scoping IAM permissions for FIS to run a specific action against a specific target; setting up Aurora failover, EKS pod disruption, ECS task stop, EC2 API error injection, network blackhole\u002Flatency\u002Floss, or Lambda async invocation experiments; or wiring experiment logging to S3 and CloudWatch Logs. Do NOT invoke for production traffic reversal, incident command, or post-incident RCA — use the appropriate troubleshoot or operate skill.\n  activation_triggers: FIS experiment template, fault injection experiment, chaos engineering, game day drill, resilience validation, Aurora failover test, RDS failover experiment, EKS pod disruption, ECS stop task experiment, EC2 stop instances experiment, EC2 send API error, network blackhole experiment, network latency injection, network packet loss, Lambda invoke async experiment, FIS stop condition, alarm-based stop\n  invocation_schema: 'Input: either (a) an experiment spec including target resource tags\u002F ARNs\u002Ffilters, one or more fault actions with parameters, optional CloudWatch alarm stop conditions, optional logging config, and budgetDuration; or (b) a partial spec for interactive refinement (e.g., \"FIS experiment to fail over an Aurora cluster\"). Output: a deterministic EXPERIMENT_TEMPLATE \u002F VERDICT \u002F CHECKLIST \u002F VERIFICATION_COMMANDS block per the STRICT output contract, where VERDICT is READY_TO_DEPLOY or PREREQUISITES_MISSING.'\n  version: 0.1.0\n  author: Jacky Chan — AWS Community Builder\n  keywords: aws, fis, fault injection, chaos engineering, experiment template, stop condition, alarm-based stop, aurora failover, rds failover, eks pod disruption, ecs stop-task, ec2 stop-instances, ec2 send-api-error, network blackhole, network latency, network loss, lambda invoke-async, game day, resilience validation, cloudops, deploy\n  tags: aws, fis, fault-injection, chaos-engineering, experiment-template, stop-condition, aurora-failover, eks-disruption, deploy, management\n  dependencies: aws-orchestrator\n---\n\n# FIS Experiment Deployer\n\n## What this skill does\n\nProvisions AWS Fault Injection Service (FIS) experiment templates with\ncorrect production-safe defaults: a least-privilege IAM execution role,\nalarm-based stop conditions, target scoping via tags\u002FARNs\u002Ffilters,\naction parameters tuned to the fault type, and experiment logging. The\nskill walks an 8-step procedure, surfaces the silent-failure modes\nunique to FIS (most dangerous: an alarm with insufficient permissions\nsilently never fires as a stop condition, so an out-of-control\nexperiment runs to budget with no automatic halt), and emits a\nREADY_TO_DEPLOY checklist verifying every action, target, and stop\ncondition against the IAM and CloudWatch state. The single most common\nincident this skill prevents: an operator points `aws:ec2:stop-instances`\nat a broad tag selector (`env=prod`), with no stop condition, in an\naccount where the FIS role has permission to stop every EC2 instance —\nthe experiment stops an entire production fleet because the resource\nfilter was permissive.\n\n## Quick navigation\n\n| Section | What it covers | When to read it |\n|---|---|---|\n| **Quick reference** | Provisioning summary (8 steps), verdict thresholds, action matrix | Before any operation |\n| **Activation keywords** | Phrases that route to this skill | Disambiguating routing |\n| **Invocation contract** | Required literal labels in the response | Formatting the response |\n| **Reasoning framework** | Why the 8-step order matters; blast-radius model | Understanding the deploy model |\n| **Dependency graph** | Which configs silently no-op without their prerequisite | Debugging \"experiment didn't run as expected\" |\n| **Expert heuristic** | Target scoping myth; stop-condition silent no-op; budgetDuration | Pre-empt production incidents |\n| **Prerequisites** | What to verify before emitting any command | Avoid PREREQUISITES_MISSING rework |\n| **8-step procedure** | The actual provisioning with copy-pasteable CLI | Executing the deploy |\n| **NEVER (top 5)** | Hard rules that prevent production outage \u002F data-plane bugs | Review before deploy |\n| **STRICT output contract** | Required EXPERIMENT_TEMPLATE \u002F VERDICT \u002F CHECKLIST \u002F VERIFICATION_COMMANDS block | Formatting the response |\n| **Recent AWS features** | Aurora failover, EKS disruption, network fault actions | Stay current |\n\n## Quick reference — provisioning summary (8 steps)\n\n| Step | Action | Reversible? | Key risk if skipped |\n|---|---|---|---|\n| 1 | Confirm experiment intent + blast radius | — | wrong intent = wrong fault |\n| 2 | Scope targets via tags \u002F ARNs \u002F filters | Yes (template only) | permissive target = production outage |\n| 3 | Choose fault action(s) + parameters | Yes | wrong action = wrong fault |\n| 4 | Configure stop conditions (CloudWatch alarms) | Yes | missing\u002Fno-permission = no auto-halt |\n| 5 | Create IAM execution role with least privilege | Yes | over-scoped role = blast-radius multiplier |\n| 6 | Configure experiment logging (S3 \u002F CloudWatch Logs) | Yes | no logging = no post-experiment evidence |\n| 7 | Set budgetDuration + (optional) schedule | Yes | long budget = long exposure |\n| 8 | Verify via `get-experiment-template` + emit checklist | — | silent no-ops |\n\n**Critical ordering constraints:** targets scoped before actions (the\ntarget filter determines blast radius; an action chosen first invites\n\"point this at everything\"); stop conditions resolved before IAM role\ncreation (the role needs `cloudwatch:DescribeAlarms` for stop-condition\nalarms); IAM role scoped to the exact actions + targets before template\ncreation (over-scoped role persists even after the experiment); logging\nconfigured before the experiment starts (FIS does not back-fill logs);\nbudgetDuration set explicitly per action (the default is the action's\nnatural duration, which may be longer than your maintenance window).\nRationale and the silent-failure table are below.\n\n## Activation keywords\n\nFIS experiment template, fault injection experiment, chaos engineering,\ngame day drill, resilience validation, Aurora failover test, RDS\nfailover experiment, EKS pod disruption, ECS stop-task experiment, EC2\nstop-instances experiment, EC2 send-api-error, EC2 terminate-instances,\nnetwork blackhole experiment, network latency injection, network packet\nloss, Lambda invoke-async experiment, FIS stop condition, alarm-based\nstop, FIS IAM role, FIS budgetDuration.\n\n## Invocation contract (hard requirement)\n\nWhen this skill is invoked with an FIS experiment provisioning request,\nthe agent MUST respond with the checklist defined in §\"STRICT output\ncontract\" using the literal all-caps labels `EXPERIMENT_TEMPLATE:`,\n`VERDICT:`, `CHECKLIST:`, and `VERIFICATION_COMMANDS:`. Do NOT preface\nwith prose, headings, or disclaimers — emit the block as the first\nlines of the response. This contract is what assertion-based evals and\ndownstream provisioning pipelines rely on; deviating from the literal\nlabels breaks automation silently.\n\n## Reasoning framework (why the provisioning order matters)\n\nFIS looks like \"pick a fault and a target\" but the underlying model has\nfour traps:\n\n1. **A fault action is a payload; the target filter is the blast\n   radius.** Operators focus on the action (`aws:ec2:stop-instances`)\n   and treat the target as boilerplate. The action does exactly what it\n   says; the target filter decides which resources it does it to. A\n   broad filter (`resourceTags: env=prod`) on a stop action with no\n   stop condition is a production outage waiting for the next\n   `start-experiment`.\n\n2. **Stop conditions are CloudWatch alarms — and the FIS role must have\n   permission to describe them.** A stop condition is an alarm ARN that\n   FIS evaluates continuously during the experiment. If the alarm goes\n   ALARM, FIS halts the experiment. But FIS needs `cloudwatch:Describe-\n   Alarms` on the alarm ARN; without it, the alarm state is invisible\n   to FIS and the experiment runs to budget regardless. This is a\n   silent no-op — the alarm exists, the stop condition is configured,\n   the role lacks the permission, and there is no error.\n\n3. **The FIS execution role is NOT the caller's role.** FIS assumes a\n   role you provide when it runs the experiment. That role's policy\n   determines what FIS can do — `ec2:StopInstances` on which instances,\n   `rds:FailoverDBCluster` on which clusters, and so on. The caller\n   being a full admin does NOT narrow the FIS role's permissions; the\n   FIS role is the only thing that matters at experiment time.\n\n4. **`budgetDuration` bounds the experiment but does NOT bound each\n   action's effect.** A 5-minute budget on an EC2 stop-instances action\n   means FIS will stop instances and then, after the budget, attempt\n   the action's rollback (typically start-instances). But if the\n   rollback fails (network partition, IAM drift), the instances stay\n   stopped. The budget is a timeout, not a guarantee of recovery.\n\nThe procedure below sequences the steps to surface these traps.\n\n## Dependency graph (silent-failure table)\n\n| Configuration | Hard dependencies (API error without) | Silent failure mode (returns 200, does nothing useful) | Enables downstream |\n|---|---|---|---|\n| Experiment template | target, action, role ARN | **template with permissive target filter = production outage on first start-experiment** | chaos drill |\n| Target (resourceTags) | tag exists on at least one resource | **tags matching many resources silently expands blast radius** | scoped fault |\n| Target (resourceArns) | ARNs resolvable | **stale ARN silently targets wrong resource after recreation** | precise fault |\n| Target (filters) | filter syntax valid | **empty filter result = experiment runs against zero resources (success, no-op)** | filtered fault |\n| Stop condition (alarm) | alarm in ALARM\u002FOK\u002FINSUFFICIENT_DATA | **alarm exists but role lacks cloudwatch:DescribeAlarms = stop condition silently inactive** | auto-halt |\n| IAM role (FIS trust) | `Service: fis.amazonaws.com` | **wrong trust principal = experiment fails at start with AccessDenied** | FIS execution |\n| IAM role (action scope) | permissions for every action's API | **missing ec2:StopInstances = action marked FAILED but experiment continues other actions** | scoped execution |\n| Logging (S3) | bucket exists + fis-acct write | **bucket policy missing fis delivery grant = logs never arrive, no error** | post-experiment evidence |\n| Logging (CloudWatch Logs) | log group exists | **log group deleted = no log streams, no error** | post-experiment evidence |\n| Aurora failover action | cluster in AVAILABLE | **cluster already failing over = action fails with ConflictException** | DR drill |\n| Network actions (SSM) | SSM agent running on target | **SSM agent offline = command never executes, no error, action \"succeeds\"** | network chaos |\n| EKS pod disruption | SSM on worker node + kubeconfig | **SSM command runs but kubectl fails silently = no pods disrupted** | pod-level chaos |\n\n**The four most dangerous silent-failure rows** are permissive-target,\nstop-condition-without-permission, network-action-SSM-offline, and\nEKS-pod-disruption-kubectl-fail. All return success on apply or even\non experiment start; the failure surfaces only when traffic doesn't\nbehave as expected. Step 2 (target scoping), Step 4 (stop-condition\nalarm verify), Step 5 (IAM permission map), and Step 8\n(`get-experiment-template` + per-action dry-run on one resource) are\nnon-negotiable for any production-adjacent experiment.\n\n## Expert heuristic: the permissive-target myth\n\nWhy targets — not actions — define blast radius, and the tag-scoping model, moved to references.\n→ [references\u002Fadvanced-patterns.md](references\u002Fadvanced-patterns.md) § Expert heuristic: the permissive-target myth\n\n## Expert heuristic: stop-condition silent no-op\n\nAlarm-permission trap and polling semantics moved to references.\n→ [references\u002Fadvanced-patterns.md](references\u002Fadvanced-patterns.md) § Expert heuristic: stop-condition silent no-op\n\n## Expert heuristic: budgetDuration vs. action natural duration\n\nPer-action natural-duration model and rollback bounds moved to references.\n→ [references\u002Fadvanced-patterns.md](references\u002Fadvanced-patterns.md) § Expert heuristic: budgetDuration vs. action natural duration\n\n## Prerequisites (verify before provisioning)\n\nBefore emitting any provisioning command, verify these prerequisites.\nIf any are missing, the verdict is **PREREQUISITES_MISSING** with a\nspecific gap citation.\n\n| Prerequisite | Why it matters | How to verify |\n|---|---|---|\n| Caller IAM | `fis:CreateExperimentTemplate`, `iam:CreateRole`, `iam:PassRole` required | `aws sts get-caller-identity` + IAM policy check |\n| Target resources exist | Empty target = silent no-op experiment | `aws ec2 describe-instances --filters Name=tag:\u003Ck>,Values=\u003Cv>` |\n| Target tag is reserved for FIS | Permissive tag (`env=prod`) = production outage | Confirm tag value population; prefer `fis-target=true` |\n| CloudWatch alarm exists (stop condition) | Stop condition ARN must be resolvable | `aws cloudwatch describe-alarms --alarm-names \u003Cname>` |\n| Alarm not in INSUFFICIENT_DATA | Stop condition silently inactive if no data | `aws cloudwatch describe-alarms --alarm-names \u003Cname> --query 'MetricAlarms[0].StateValue'` |\n| SSM agent reachable (network actions) | Network actions run via SSM; offline agent = no-op | `aws ssm describe-instance-information --filters Name=InstanceIds,Values=\u003Cid>` |\n| Aurora cluster AVAILABLE (failover) | Cluster mid-failover = action fails | `aws rds describe-db-clusters --db-cluster-identifier \u003Cid>` |\n| EKS worker node has SSM + kubectl (EKS disruption) | SSM command runs but kubectl fails silently otherwise | `aws ssm describe-instance-information` + `aws eks describe-cluster` |\n\nIf any prerequisite is missing, output `VERDICT: PREREQUISITES_MISSING`\nand cite the specific gap.\n\n## 8-step provisioning procedure\n\n### Step 1 — Confirm experiment intent + blast radius\n\nBefore any resource lookup, name the intent in one sentence: \"drain\n10% of an ASG for 60 seconds to test alerting,\" \"fail over Aurora\ncluster X to validate RTO,\" \"blackhole traffic from app-A to db-B for\n5 minutes.\" The intent drives the target filter, the action, and the\nstop condition. If the intent is \"see what breaks,\" the verdict is\nPREREQUISITES_MISSING — FIS is not a discovery tool.\n\n### Step 2 — Scope targets via tags \u002F ARNs \u002F filters\n\n| Target type | When to use | Verification |\n|---|---|---|\n| `resourceTags` | Drills across a labeled group (e.g., fis-target=true) | `aws ec2 describe-instances --filters` |\n| `resourceArns` | Precise single-resource faults | resource ARN resolvable |\n| `filters` (resource type + boolean) | Complex scoping (e.g., instances in ASG X with tag Y) | List + count before template |\n\n**Common mistake:** using `env=prod` or `service=critical` as a target\ntag. Always prefer a tag whose only population is FIS-eligible\nresources (`fis-target=true`). Confirm the population size before\napplying the template — a 1000-instance match on a stop action is a\nfleet-wide outage.\n\n### Step 3 — Choose fault action(s) + parameters\n\n| Action ID | Effect | Key parameters | Natural duration |\n|---|---|---|---|\n| `aws:ec2:stop-instances` | Stops the matched EC2 instances | duration (PTxS) | stops, then starts at action end |\n| `aws:ec2:send-api-error` | Returns error for IAM-role API calls | errorCode (ThrottlingException, etc.), service, duration | effect ends with action |\n| `aws:ec2:terminate-instances` | Terminates matched EC2 instances | (none — irreversible) | irreversible |\n| `aws:ecs:stop-task` | Stops matched ECS tasks | (none) | task stopped, ECS reschedules |\n| `aws:eks:pod-cleanup` (SSM) | Deletes pods via kubectl on worker node | kubectl command via SSM | pods deleted, controller reschedules |\n| `aws:network:disrupt-connectivity` (SSM) | Network blackhole\u002Flatency\u002Floss via tc | networkInterfaces, disruptionType, duration | rules persist if SSM offline at end |\n| `aws:rds:failover-db-cluster` | Promotes a different Aurora writer | (none) | 30-120s failover |\n| `aws:lambda:invoke-async` | Invokes Lambda with payload asynchronously | function ARN, payload | invocation completes |\n\n**Common mistake:** using `terminate-instances` for a chaos drill when\nyou meant `stop-instances`. Terminate is irreversible — the ASG will\nprovision replacements, but state is lost. Always use stop for\nstateful drills.\n\n### Step 4 — Configure stop conditions (CloudWatch alarms)\n\nA stop condition is an array of CloudWatch alarm ARNs. FIS evaluates\nthem continuously; any alarm going ALARM halts the experiment.\n\n```bash\naws cloudwatch describe-alarms --alarm-names \u003CALARM_NAME> \\\n  --query 'MetricAlarms[0].{Name:AlarmName,State:StateValue,Arn:AlarmArn}' \\\n  --output json\n```\n\nConfirm:\n1. The alarm exists (a typo'd name = silent no-op stop condition).\n2. The alarm is in OK or ALARM state (not INSUFFICIENT_DATA — that\n   means it will never fire as a stop condition).\n3. The FIS role has `cloudwatch:DescribeAlarms` on the alarm ARN.\n\n### Step 5 — Create IAM execution role with least privilege\n\nFull least-privilege trust policy, permissions policy templates, and condition scoping moved to references.\n→ [references\u002Fiam-and-logging-templates.md](references\u002Fiam-and-logging-templates.md) § Step 5 — Create IAM execution role\n\n### Step 6 — Configure experiment logging (S3 \u002F CloudWatch Logs)\n\nS3 \u002F CloudWatch Logs logConfiguration JSON templates and prerequisites moved to references.\n→ [references\u002Fiam-and-logging-templates.md](references\u002Fiam-and-logging-templates.md) § Step 6 — Configure experiment logging\n\n### Step 7 — Set budgetDuration + (optional) schedule\n\n```json\n\"budgetDuration\": \"PT5M\"\n```\n\nSet 2x the action's expected effect duration. For multi-action\nexperiments, the budget bounds the total — set it to the sum of\nsequential action durations plus 50% margin.\n\nFor scheduled game days, use a CloudWatch Events rule that invokes\n`fis:start-experiment` at the scheduled time. Do NOT leave an\nauto-start rule armed in production — disarm between drills.\n\n### Step 8 — Verify via `get-experiment-template` + emit checklist\n\n```bash\naws fis get-experiment-template --id \u003CTEMPLATE_ID>\n\n# Optional: dry-run by starting the experiment and aborting within 30s\n# aws fis start-experiment --experiment-template-id \u003CTEMPLATE_ID>\n# aws fis stop-experiment --id \u003CEXPERIMENT_ID>\n```\n\nConfirm every action ID, every target filter, every stop condition\nalarm ARN, and the IAM role ARN are present in the stored template.\nFIS does not validate IAM permissions at template creation time — only\nat experiment start, where the first action fails with AccessDenied.\n\n## NEVER do these things\n\nThese anti-patterns cause production outages, data-plane bugs, or\nsilent no-ops. Each is observed in real production incidents.\n\n1. **NEVER scope an FIS target with a broad production tag\n   (`env=prod`, `service=critical`).** Why it's wrong: the target\n   filter IS the blast radius. A stop action against `env=prod` stops\n   every matching instance. Use a tag reserved for FIS\n   (`fis-target=true`), and scope the IAM role with the same tag as a\n   second line of defense.\n\n2. **NEVER deploy an experiment template without at least one stop\n   condition.** Why it's wrong: without an alarm, an out-of-control\n   experiment runs to budgetDuration with no automatic halt. Even for\n   \"safe\" drills, configure at least an \"error rate above threshold\"\n   alarm. A no-stop-condition experiment is a manual kill or wait-\n   for-budget experiment.\n\n3. **NEVER grant the FIS execution role `Resource: \"*\"` on the action\n   APIs without tag conditions.** Why it's wrong: the role is the\n   ONLY thing that scopes the experiment at runtime. An admin caller\n   with a permissive FIS role still produces a fleet-wide outage. The\n   tag condition is the second line of defense when the target filter\n   is wrong.\n\n4. **NEVER assume a stop condition will fire without verifying\n   `cloudwatch:DescribeAlarms` on the alarm ARN in the FIS role.** Why\n   it's wrong: FIS fails open on missing alarm permission — the\n   experiment runs to budget with the stop condition silently\n   inactive. Verify by reading the role policy and by force-triggering\n   the alarm in a dry run.\n\n5. **NEVER use `aws:ec2:terminate-instances` when you mean\n   `stop-instances`.** Why it's wrong: terminate is irreversible. The\n   ASG provisions replacements, but instance state and ephemeral disks\n   are gone. For chaos drills where the workload should recover, use\n   stop-instances. Reserve terminate-instances for tests of stateless\n   auto-replacement, never stateful services.\n\n## Output format\n\n```\nEXPERIMENT_TEMPLATE: \u003Cname> (action: \u003Caction-id>, target: \u003Ctarget-summary>)\nVERDICT: READY_TO_DEPLOY | PREREQUISITES_MISSING\nCHECKLIST:\n  [✓|✗] Intent + blast radius: \u003Cone-sentence intent>, blast radius \u003Ccount> \u003Ctype>\n  [✓|✗] Target scoped: \u003CresourceTags|resourceArns|filters> (\u003Ccount> resources)\n  [✓|✗] Action(s): \u003Caction-id list> (parameters: \u003Ckey params>)\n  [✓|✗] Stop condition(s): \u003Calarm-name list> (state: OK|ALARM)\n  [✓|✗] IAM role: \u003Crole-name> (fis trust, least-privilege actions, tag conditions)\n  [✓|✗] Logging: S3 \u003Cbucket> + CloudWatch Logs \u003Clog-group>\n  [✓|✗] budgetDuration: \u003CPTxM> (rationale: \u003Caction-appropriate>)\nVERIFICATION_COMMANDS:\n  \u003Ccopy-pasteable verification commands>\n```\n\n## STRICT output contract\n\nThe rules below are hard constraints. Violating any one produces a\nchecklist that looks complete but contains a silent misconfiguration.\nSelf-check EVERY emitted block against these rules before returning.\n\n### Required output structure\n\nEvery response MUST be a single block using these literal labels, in\nthis order. Do NOT preface with prose, headings, or disclaimers.\n\n```text\nEXPERIMENT_TEMPLATE: \u003Cname> (action: \u003Caction-id>, target: \u003Ctarget-summary>)\nVERDICT: READY_TO_DEPLOY | PREREQUISITES_MISSING\nCHECKLIST:\n  [✓|✗] Intent + blast radius: \u003Cone-sentence intent>, blast radius \u003Ccount> \u003Ctype>\n  [✓|✗] Target scoped: \u003CresourceTags|resourceArns|filters> (\u003Ccount> resources)\n  [✓|✗] Action(s): \u003Caction-id list> (parameters: \u003Ckey params>)\n  [✓|✗] Stop condition(s): \u003Calarm-name list> (state: OK|ALARM)\n  [✓|✗] IAM role: \u003Crole-name> (fis trust, least-privilege actions, tag conditions)\n  [✓|✗] Logging: S3 \u003Cbucket> + CloudWatch Logs \u003Clog-group>\n  [✓|✗] budgetDuration: \u003CPTxM> (rationale: \u003Caction-appropriate>)\nVERIFICATION_COMMANDS:\n  \u003Ccopy-pasteable verification commands — one per [✓] item>\n```\n\n### FORBIDDEN output patterns\n\n1. **NEVER emit `VERDICT: READY_TO_DEPLOY` without showing ALL 7\n   checklist items.** Every item MUST appear with a status marker:\n   `[✓]`, `[✗]`, or `[OPTIONAL]`. Omitting a row implies it was not\n   evaluated.\n\n2. **NEVER mark Target scoped `[✓] resourceTags` without citing the\n   resource count.** A permissive tag is the leading cause of FIS\n   outages. The checklist MUST cite the resolved count\n   (e.g., \"3 instances\") or mark the item `[✗]` with a warning.\n\n3. **NEVER mark Stop condition(s) `[✓]` without confirming the alarm\n   state is OK or ALARM (not INSUFFICIENT_DATA) AND the FIS role has\n   `cloudwatch:DescribeAlarms` on the ARN.** Both checks are required;\n   either missing means the stop condition is silently inactive. The\n   checklist MUST cite the alarm state and the role's alarm permission.\n\n4. **NEVER mark IAM role `[✓]` without citing that the action\n   permissions carry tag conditions** (e.g.,\n   `aws:ResourceTag\u002Ffis-target=true`). Without the condition, the role\n   is the only line of defense — and over-scoped roles are the second-\n   most-common FIS incident.\n\n5. **NEVER mark budgetDuration `[✓]` without a one-line rationale\n   tied to the action.** A 30-minute budget on a stop-instances drill\n   is 30 minutes of stopped production; a 30-second budget on a\n   network blackhole may be too short for the SSM command to land.\n   The checklist MUST include the rationale.\n\n6. **NEVER emit `VERDICT: PREREQUISITES_MISSING` without citing the\n   specific gap.** Each `[✗]` item MUST have a one-line reason:\n   `[✗] Stop condition: alarm fis-stop-rate not found — create alarm\n   before applying`. A bare `[✗]` is non-compliant.\n\n7. **NEVER include `aws:ec2:terminate-instances` in an action list\n   without a `[WARN]` note in the checklist.** Terminate is\n   irreversible. The checklist MUST cite that stateful resources will\n   be lost.\n\n### Perfect example output — READY_TO_DEPLOY\n\n```text\nEXPERIMENT_TEMPLATE: ec2-stop-canary (action: aws:ec2:stop-instances, target: fis-target=true)\nVERDICT: READY_TO_DEPLOY\nCHECKLIST:\n  [✓] Intent + blast radius: Stop one canary instance for 60s to validate auto-recovery, blast radius 1 instance\n  [✓] Target scoped: resourceTags fis-target=true (1 instance i-0abc123)\n  [✓] Action(s): aws:ec2:stop-instances (duration PT60S)\n  [✓] Stop condition(s): fis-stop-error-rate (state: OK)\n  [✓] IAM role: fis-execution-role (fis trust, ec2:StopInstances + StartInstances scoped to aws:ResourceTag\u002Ffis-target=true, cloudwatch:DescribeAlarms on fis-stop-error-rate)\n  [✓] Logging: S3 fis-logs-111111111111 + CloudWatch Logs \u002Faws\u002Ffis\u002Fec2-stop-canary\n  [✓] budgetDuration: PT2M (rationale: 2x the 60s stop action for rollback margin)\nVERIFICATION_COMMANDS:\n  aws fis get-experiment-template --id EXT-AAAA1111\n  aws ec2 describe-instances --filters Name=tag:fis-target,Values=true --query 'Reservations[0].Instances[0].InstanceId'\n  aws cloudwatch describe-alarms --alarm-names fis-stop-error-rate --query 'MetricAlarms[0].StateValue'\n  aws iam get-role-policy --role-name fis-execution-role --policy-name fis-execution-policy\n  aws logs describe-log-groups --log-group-name-prefix \u002Faws\u002Ffis\u002Fec2-stop-canary\n```\n\n### Perfect example output — PREREQUISITES_MISSING\n\nFull PREREQUISITES_MISSING example output moved to references.\n→ [references\u002Fworked-examples.md](references\u002Fworked-examples.md) § Perfect example output — PREREQUISITES_MISSING\n\n## Recent AWS features\n\n2024-2026 FIS feature notes (Aurora failover, EKS actions, SSM-based network disruption, org-ID conditions) moved to references.\n→ [references\u002Fadvanced-patterns.md](references\u002Fadvanced-patterns.md) § Recent AWS features\n\n## AWS documentation\n\n- **AWS Fault Injection Service User Guide** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002F\n- **FIS Actions Reference** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002Ffis-actions-reference.html\n- **FIS Target Types** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002Ftargets.html\n- **FIS Stop Conditions** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002Fstop-conditions.html\n- **FIS IAM Roles** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002Fiam-service-role.html\n- **FIS Logging** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002Flogging.html\n- **FIS API Reference** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002FAPIReference\u002F\n- **FIS CLI Reference** — https:\u002F\u002Fdocs.aws.amazon.com\u002Fcli\u002Flatest\u002Freference\u002Ffis\u002F\n- **Blog: AWS Resilience Hub + FIS** — https:\u002F\u002Faws.amazon.com\u002Fblogs\u002Fmt\u002Ftest-application-resilience-using-aws-resilience-hub-and-aws-fault-injection-service\u002F\n- **Workshop: chaos engineering with FIS** — https:\u002F\u002Fcatalog.workshops.aws\u002Fchaosengineering\n\n## References (load on demand)\n\nConsult these only when the corresponding topic comes up:\n\n- [references\u002Ffault-action-catalog.md](references\u002Ffault-action-catalog.md) — fault action IDs, parameters, and rollback semantics\n- [references\u002Fiam-and-logging-templates.md](references\u002Fiam-and-logging-templates.md) — execution-role policies (moved from Step 5) and logging config (moved from Step 6)\n- [references\u002Fworked-examples.md](references\u002Fworked-examples.md) — the PREREQUISITES_MISSING full-output example (moved from § Output format)\n- [references\u002Fadvanced-patterns.md](references\u002Fadvanced-patterns.md) — expert heuristics (permissive-target myth, stop-condition no-op, budgetDuration) and recent AWS features\n","# FIS Experiment Deployer\n\n## What this skill does\n\nProvisions AWS Fault Injection Service (FIS) experiment templates with\ncorrect production-safe defaults: a least-privilege IAM execution role,\nalarm-based stop conditions, target scoping via tags\u002FARNs\u002Ffilters,\naction parameters tuned to the fault type, and experiment logging. The\nskill walks an 8-step procedure, surfaces the silent-failure modes\nunique to FIS (most dangerous: an alarm with insufficient permissions\nsilently never fires as a stop condition, so an out-of-control\nexperiment runs to budget with no automatic halt), and emits a\nREADY_TO_DEPLOY checklist verifying every action, target, and stop\ncondition against the IAM and CloudWatch state. The single most common\nincident this skill prevents: an operator points `aws:ec2:stop-instances`\nat a broad tag selector (`env=prod`), with no stop condition, in an\naccount where the FIS role has permission to stop every EC2 instance —\nthe experiment stops an entire production fleet because the resource\nfilter was permissive.\n\n## Quick navigation\n\n| Section | What it covers | When to read it |\n|---|---|---|\n| **Quick reference** | Provisioning summary (8 steps), verdict thresholds, action matrix | Before any operation |\n| **Activation keywords** | Phrases that route to this skill | Disambiguating routing |\n| **Invocation contract** | Required literal labels in the response | Formatting the response |\n| **Reasoning framework** | Why the 8-step order matters; blast-radius model | Understanding the deploy model |\n| **Dependency graph** | Which configs silently no-op without their prerequisite | Debugging \"experiment didn't run as expected\" |\n| **Expert heuristic** | Target scoping myth; stop-condition silent no-op; budgetDuration | Pre-empt production incidents |\n| **Prerequisites** | What to verify before emitting any command | Avoid PREREQUISITES_MISSING rework |\n| **8-step procedure** | The actual provisioning with copy-pasteable CLI | Executing the deploy |\n| **NEVER (top 5)** | Hard rules that prevent production outage \u002F data-plane bugs | Review before deploy |\n| **STRICT output contract** | Required EXPERIMENT_TEMPLATE \u002F VERDICT \u002F CHECKLIST \u002F VERIFICATION_COMMANDS block | Formatting the response |\n| **Recent AWS features** | Aurora failover, EKS disruption, network fault actions | Stay current |\n\n## Quick reference — provisioning summary (8 steps)\n\n| Step | Action | Reversible? | Key risk if skipped |\n|---|---|---|---|\n| 1 | Confirm experiment intent + blast radius | — | wrong intent = wrong fault |\n| 2 | Scope targets via tags \u002F ARNs \u002F filters | Yes (template only) | permissive target = production outage |\n| 3 | Choose fault action(s) + parameters | Yes | wrong action = wrong fault |\n| 4 | Configure stop conditions (CloudWatch alarms) | Yes | missing\u002Fno-permission = no auto-halt |\n| 5 | Create IAM execution role with least privilege | Yes | over-scoped role = blast-radius multiplier |\n| 6 | Configure experiment logging (S3 \u002F CloudWatch Logs) | Yes | no logging = no post-experiment evidence |\n| 7 | Set budgetDuration + (optional) schedule | Yes | long budget = long exposure |\n| 8 | Verify via `get-experiment-template` + emit checklist | — | silent no-ops |\n\n**Critical ordering constraints:** targets scoped before actions (the\ntarget filter determines blast radius; an action chosen first invites\n\"point this at everything\"); stop conditions resolved before IAM role\ncreation (the role needs `cloudwatch:DescribeAlarms` for stop-condition\nalarms); IAM role scoped to the exact actions + targets before template\ncreation (over-scoped role persists even after the experiment); logging\nconfigured before the experiment starts (FIS does not back-fill logs);\nbudgetDuration set explicitly per action (the default is the action's\nnatural duration, which may be longer than your maintenance window).\nRationale and the silent-failure table are below.\n\n## Activation keywords\n\nFIS experiment template, fault injection experiment, chaos engineering,\ngame day drill, resilience validation, Aurora failover test, RDS\nfailover experiment, EKS pod disruption, ECS stop-task experiment, EC2\nstop-instances experiment, EC2 send-api-error, EC2 terminate-instances,\nnetwork blackhole experiment, network latency injection, network packet\nloss, Lambda invoke-async experiment, FIS stop condition, alarm-based\nstop, FIS IAM role, FIS budgetDuration.\n\n## Invocation contract (hard requirement)\n\nWhen this skill is invoked with an FIS experiment provisioning request,\nthe agent MUST respond with the checklist defined in §\"STRICT output\ncontract\" using the literal all-caps labels `EXPERIMENT_TEMPLATE:`,\n`VERDICT:`, `CHECKLIST:`, and `VERIFICATION_COMMANDS:`. Do NOT preface\nwith prose, headings, or disclaimers — emit the block as the first\nlines of the response. This contract is what assertion-based evals and\ndownstream provisioning pipelines rely on; deviating from the literal\nlabels breaks automation silently.\n\n## Reasoning framework (why the provisioning order matters)\n\nFIS looks like \"pick a fault and a target\" but the underlying model has\nfour traps:\n\n1. **A fault action is a payload; the target filter is the blast\n   radius.** Operators focus on the action (`aws:ec2:stop-instances`)\n   and treat the target as boilerplate. The action does exactly what it\n   says; the target filter decides which resources it does it to. A\n   broad filter (`resourceTags: env=prod`) on a stop action with no\n   stop condition is a production outage waiting for the next\n   `start-experiment`.\n\n2. **Stop conditions are CloudWatch alarms — and the FIS role must have\n   permission to describe them.** A stop condition is an alarm ARN that\n   FIS evaluates continuously during the experiment. If the alarm goes\n   ALARM, FIS halts the experiment. But FIS needs `cloudwatch:Describe-\n   Alarms` on the alarm ARN; without it, the alarm state is invisible\n   to FIS and the experiment runs to budget regardless. This is a\n   silent no-op — the alarm exists, the stop condition is configured,\n   the role lacks the permission, and there is no error.\n\n3. **The FIS execution role is NOT the caller's role.** FIS assumes a\n   role you provide when it runs the experiment. That role's policy\n   determines what FIS can do — `ec2:StopInstances` on which instances,\n   `rds:FailoverDBCluster` on which clusters, and so on. The caller\n   being a full admin does NOT narrow the FIS role's permissions; the\n   FIS role is the only thing that matters at experiment time.\n\n4. **`budgetDuration` bounds the experiment but does NOT bound each\n   action's effect.** A 5-minute budget on an EC2 stop-instances action\n   means FIS will stop instances and then, after the budget, attempt\n   the action's rollback (typically start-instances). But if the\n   rollback fails (network partition, IAM drift), the instances stay\n   stopped. The budget is a timeout, not a guarantee of recovery.\n\nThe procedure below sequences the steps to surface these traps.\n\n## Dependency graph (silent-failure table)\n\n| Configuration | Hard dependencies (API error without) | Silent failure mode (returns 200, does nothing useful) | Enables downstream |\n|---|---|---|---|\n| Experiment template | target, action, role ARN | **template with permissive target filter = production outage on first start-experiment** | chaos drill |\n| Target (resourceTags) | tag exists on at least one resource | **tags matching many resources silently expands blast radius** | scoped fault |\n| Target (resourceArns) | ARNs resolvable | **stale ARN silently targets wrong resource after recreation** | precise fault |\n| Target (filters) | filter syntax valid | **empty filter result = experiment runs against zero resources (success, no-op)** | filtered fault |\n| Stop condition (alarm) | alarm in ALARM\u002FOK\u002FINSUFFICIENT_DATA | **alarm exists but role lacks cloudwatch:DescribeAlarms = stop condition silently inactive** | auto-halt |\n| IAM role (FIS trust) | `Service: fis.amazonaws.com` | **wrong trust principal = experiment fails at start with AccessDenied** | FIS execution |\n| IAM role (action scope) | permissions for every action's API | **missing ec2:StopInstances = action marked FAILED but experiment continues other actions** | scoped execution |\n| Logging (S3) | bucket exists + fis-acct write | **bucket policy missing fis delivery grant = logs never arrive, no error** | post-experiment evidence |\n| Logging (CloudWatch Logs) | log group exists | **log group deleted = no log streams, no error** | post-experiment evidence |\n| Aurora failover action | cluster in AVAILABLE | **cluster already failing over = action fails with ConflictException** | DR drill |\n| Network actions (SSM) | SSM agent running on target | **SSM agent offline = command never executes, no error, action \"succeeds\"** | network chaos |\n| EKS pod disruption | SSM on worker node + kubeconfig | **SSM command runs but kubectl fails silently = no pods disrupted** | pod-level chaos |\n\n**The four most dangerous silent-failure rows** are permissive-target,\nstop-condition-without-permission, network-action-SSM-offline, and\nEKS-pod-disruption-kubectl-fail. All return success on apply or even\non experiment start; the failure surfaces only when traffic doesn't\nbehave as expected. Step 2 (target scoping), Step 4 (stop-condition\nalarm verify), Step 5 (IAM permission map), and Step 8\n(`get-experiment-template` + per-action dry-run on one resource) are\nnon-negotiable for any production-adjacent experiment.\n\n## Expert heuristic: the permissive-target myth\n\nWhy targets — not actions — define blast radius, and the tag-scoping model, moved to references.\n→ [references\u002Fadvanced-patterns.md](references\u002Fadvanced-patterns.md) § Expert heuristic: the permissive-target myth\n\n## Expert heuristic: stop-condition silent no-op\n\nAlarm-permission trap and polling semantics moved to references.\n→ [references\u002Fadvanced-patterns.md](references\u002Fadvanced-patterns.md) § Expert heuristic: stop-condition silent no-op\n\n## Expert heuristic: budgetDuration vs. action natural duration\n\nPer-action natural-duration model and rollback bounds moved to references.\n→ [references\u002Fadvanced-patterns.md](references\u002Fadvanced-patterns.md) § Expert heuristic: budgetDuration vs. action natural duration\n\n## Prerequisites (verify before provisioning)\n\nBefore emitting any provisioning command, verify these prerequisites.\nIf any are missing, the verdict is **PREREQUISITES_MISSING** with a\nspecific gap citation.\n\n| Prerequisite | Why it matters | How to verify |\n|---|---|---|\n| Caller IAM | `fis:CreateExperimentTemplate`, `iam:CreateRole`, `iam:PassRole` required | `aws sts get-caller-identity` + IAM policy check |\n| Target resources exist | Empty target = silent no-op experiment | `aws ec2 describe-instances --filters Name=tag:\u003Ck>,Values=\u003Cv>` |\n| Target tag is reserved for FIS | Permissive tag (`env=prod`) = production outage | Confirm tag value population; prefer `fis-target=true` |\n| CloudWatch alarm exists (stop condition) | Stop condition ARN must be resolvable | `aws cloudwatch describe-alarms --alarm-names \u003Cname>` |\n| Alarm not in INSUFFICIENT_DATA | Stop condition silently inactive if no data | `aws cloudwatch describe-alarms --alarm-names \u003Cname> --query 'MetricAlarms[0].StateValue'` |\n| SSM agent reachable (network actions) | Network actions run via SSM; offline agent = no-op | `aws ssm describe-instance-information --filters Name=InstanceIds,Values=\u003Cid>` |\n| Aurora cluster AVAILABLE (failover) | Cluster mid-failover = action fails | `aws rds describe-db-clusters --db-cluster-identifier \u003Cid>` |\n| EKS worker node has SSM + kubectl (EKS disruption) | SSM command runs but kubectl fails silently otherwise | `aws ssm describe-instance-information` + `aws eks describe-cluster` |\n\nIf any prerequisite is missing, output `VERDICT: PREREQUISITES_MISSING`\nand cite the specific gap.\n\n## 8-step provisioning procedure\n\n### Step 1 — Confirm experiment intent + blast radius\n\nBefore any resource lookup, name the intent in one sentence: \"drain\n10% of an ASG for 60 seconds to test alerting,\" \"fail over Aurora\ncluster X to validate RTO,\" \"blackhole traffic from app-A to db-B for\n5 minutes.\" The intent drives the target filter, the action, and the\nstop condition. If the intent is \"see what breaks,\" the verdict is\nPREREQUISITES_MISSING — FIS is not a discovery tool.\n\n### Step 2 — Scope targets via tags \u002F ARNs \u002F filters\n\n| Target type | When to use | Verification |\n|---|---|---|\n| `resourceTags` | Drills across a labeled group (e.g., fis-target=true) | `aws ec2 describe-instances --filters` |\n| `resourceArns` | Precise single-resource faults | resource ARN resolvable |\n| `filters` (resource type + boolean) | Complex scoping (e.g., instances in ASG X with tag Y) | List + count before template |\n\n**Common mistake:** using `env=prod` or `service=critical` as a target\ntag. Always prefer a tag whose only population is FIS-eligible\nresources (`fis-target=true`). Confirm the population size before\napplying the template — a 1000-instance match on a stop action is a\nfleet-wide outage.\n\n### Step 3 — Choose fault action(s) + parameters\n\n| Action ID | Effect | Key parameters | Natural duration |\n|---|---|---|---|\n| `aws:ec2:stop-instances` | Stops the matched EC2 instances | duration (PTxS) | stops, then starts at action end |\n| `aws:ec2:send-api-error` | Returns error for IAM-role API calls | errorCode (ThrottlingException, etc.), service, duration | effect ends with action |\n| `aws:ec2:terminate-instances` | Terminates matched EC2 instances | (none — irreversible) | irreversible |\n| `aws:ecs:stop-task` | Stops matched ECS tasks | (none) | task stopped, ECS reschedules |\n| `aws:eks:pod-cleanup` (SSM) | Deletes pods via kubectl on worker node | kubectl command via SSM | pods deleted, controller reschedules |\n| `aws:network:disrupt-connectivity` (SSM) | Network blackhole\u002Flatency\u002Floss via tc | networkInterfaces, disruptionType, duration | rules persist if SSM offline at end |\n| `aws:rds:failover-db-cluster` | Promotes a different Aurora writer | (none) | 30-120s failover |\n| `aws:lambda:invoke-async` | Invokes Lambda with payload asynchronously | function ARN, payload | invocation completes |\n\n**Common mistake:** using `terminate-instances` for a chaos drill when\nyou meant `stop-instances`. Terminate is irreversible — the ASG will\nprovision replacements, but state is lost. Always use stop for\nstateful drills.\n\n### Step 4 — Configure stop conditions (CloudWatch alarms)\n\nA stop condition is an array of CloudWatch alarm ARNs. FIS evaluates\nthem continuously; any alarm going ALARM halts the experiment.\n\n```bash\naws cloudwatch describe-alarms --alarm-names \u003CALARM_NAME> \\\n  --query 'MetricAlarms[0].{Name:AlarmName,State:StateValue,Arn:AlarmArn}' \\\n  --output json\n```\n\nConfirm:\n1. The alarm exists (a typo'd name = silent no-op stop condition).\n2. The alarm is in OK or ALARM state (not INSUFFICIENT_DATA — that\n   means it will never fire as a stop condition).\n3. The FIS role has `cloudwatch:DescribeAlarms` on the alarm ARN.\n\n### Step 5 — Create IAM execution role with least privilege\n\nFull least-privilege trust policy, permissions policy templates, and condition scoping moved to references.\n→ [references\u002Fiam-and-logging-templates.md](references\u002Fiam-and-logging-templates.md) § Step 5 — Create IAM execution role\n\n### Step 6 — Configure experiment logging (S3 \u002F CloudWatch Logs)\n\nS3 \u002F CloudWatch Logs logConfiguration JSON templates and prerequisites moved to references.\n→ [references\u002Fiam-and-logging-templates.md](references\u002Fiam-and-logging-templates.md) § Step 6 — Configure experiment logging\n\n### Step 7 — Set budgetDuration + (optional) schedule\n\n```json\n\"budgetDuration\": \"PT5M\"\n```\n\nSet 2x the action's expected effect duration. For multi-action\nexperiments, the budget bounds the total — set it to the sum of\nsequential action durations plus 50% margin.\n\nFor scheduled game days, use a CloudWatch Events rule that invokes\n`fis:start-experiment` at the scheduled time. Do NOT leave an\nauto-start rule armed in production — disarm between drills.\n\n### Step 8 — Verify via `get-experiment-template` + emit checklist\n\n```bash\naws fis get-experiment-template --id \u003CTEMPLATE_ID>\n\n# Optional: dry-run by starting the experiment and aborting within 30s\n# aws fis start-experiment --experiment-template-id \u003CTEMPLATE_ID>\n# aws fis stop-experiment --id \u003CEXPERIMENT_ID>\n```\n\nConfirm every action ID, every target filter, every stop condition\nalarm ARN, and the IAM role ARN are present in the stored template.\nFIS does not validate IAM permissions at template creation time — only\nat experiment start, where the first action fails with AccessDenied.\n\n## NEVER do these things\n\nThese anti-patterns cause production outages, data-plane bugs, or\nsilent no-ops. Each is observed in real production incidents.\n\n1. **NEVER scope an FIS target with a broad production tag\n   (`env=prod`, `service=critical`).** Why it's wrong: the target\n   filter IS the blast radius. A stop action against `env=prod` stops\n   every matching instance. Use a tag reserved for FIS\n   (`fis-target=true`), and scope the IAM role with the same tag as a\n   second line of defense.\n\n2. **NEVER deploy an experiment template without at least one stop\n   condition.** Why it's wrong: without an alarm, an out-of-control\n   experiment runs to budgetDuration with no automatic halt. Even for\n   \"safe\" drills, configure at least an \"error rate above threshold\"\n   alarm. A no-stop-condition experiment is a manual kill or wait-\n   for-budget experiment.\n\n3. **NEVER grant the FIS execution role `Resource: \"*\"` on the action\n   APIs without tag conditions.** Why it's wrong: the role is the\n   ONLY thing that scopes the experiment at runtime. An admin caller\n   with a permissive FIS role still produces a fleet-wide outage. The\n   tag condition is the second line of defense when the target filter\n   is wrong.\n\n4. **NEVER assume a stop condition will fire without verifying\n   `cloudwatch:DescribeAlarms` on the alarm ARN in the FIS role.** Why\n   it's wrong: FIS fails open on missing alarm permission — the\n   experiment runs to budget with the stop condition silently\n   inactive. Verify by reading the role policy and by force-triggering\n   the alarm in a dry run.\n\n5. **NEVER use `aws:ec2:terminate-instances` when you mean\n   `stop-instances`.** Why it's wrong: terminate is irreversible. The\n   ASG provisions replacements, but instance state and ephemeral disks\n   are gone. For chaos drills where the workload should recover, use\n   stop-instances. Reserve terminate-instances for tests of stateless\n   auto-replacement, never stateful services.\n\n## Output format\n\n```\nEXPERIMENT_TEMPLATE: \u003Cname> (action: \u003Caction-id>, target: \u003Ctarget-summary>)\nVERDICT: READY_TO_DEPLOY | PREREQUISITES_MISSING\nCHECKLIST:\n  [✓|✗] Intent + blast radius: \u003Cone-sentence intent>, blast radius \u003Ccount> \u003Ctype>\n  [✓|✗] Target scoped: \u003CresourceTags|resourceArns|filters> (\u003Ccount> resources)\n  [✓|✗] Action(s): \u003Caction-id list> (parameters: \u003Ckey params>)\n  [✓|✗] Stop condition(s): \u003Calarm-name list> (state: OK|ALARM)\n  [✓|✗] IAM role: \u003Crole-name> (fis trust, least-privilege actions, tag conditions)\n  [✓|✗] Logging: S3 \u003Cbucket> + CloudWatch Logs \u003Clog-group>\n  [✓|✗] budgetDuration: \u003CPTxM> (rationale: \u003Caction-appropriate>)\nVERIFICATION_COMMANDS:\n  \u003Ccopy-pasteable verification commands>\n```\n\n## STRICT output contract\n\nThe rules below are hard constraints. Violating any one produces a\nchecklist that looks complete but contains a silent misconfiguration.\nSelf-check EVERY emitted block against these rules before returning.\n\n### Required output structure\n\nEvery response MUST be a single block using these literal labels, in\nthis order. Do NOT preface with prose, headings, or disclaimers.\n\n```text\nEXPERIMENT_TEMPLATE: \u003Cname> (action: \u003Caction-id>, target: \u003Ctarget-summary>)\nVERDICT: READY_TO_DEPLOY | PREREQUISITES_MISSING\nCHECKLIST:\n  [✓|✗] Intent + blast radius: \u003Cone-sentence intent>, blast radius \u003Ccount> \u003Ctype>\n  [✓|✗] Target scoped: \u003CresourceTags|resourceArns|filters> (\u003Ccount> resources)\n  [✓|✗] Action(s): \u003Caction-id list> (parameters: \u003Ckey params>)\n  [✓|✗] Stop condition(s): \u003Calarm-name list> (state: OK|ALARM)\n  [✓|✗] IAM role: \u003Crole-name> (fis trust, least-privilege actions, tag conditions)\n  [✓|✗] Logging: S3 \u003Cbucket> + CloudWatch Logs \u003Clog-group>\n  [✓|✗] budgetDuration: \u003CPTxM> (rationale: \u003Caction-appropriate>)\nVERIFICATION_COMMANDS:\n  \u003Ccopy-pasteable verification commands — one per [✓] item>\n```\n\n### FORBIDDEN output patterns\n\n1. **NEVER emit `VERDICT: READY_TO_DEPLOY` without showing ALL 7\n   checklist items.** Every item MUST appear with a status marker:\n   `[✓]`, `[✗]`, or `[OPTIONAL]`. Omitting a row implies it was not\n   evaluated.\n\n2. **NEVER mark Target scoped `[✓] resourceTags` without citing the\n   resource count.** A permissive tag is the leading cause of FIS\n   outages. The checklist MUST cite the resolved count\n   (e.g., \"3 instances\") or mark the item `[✗]` with a warning.\n\n3. **NEVER mark Stop condition(s) `[✓]` without confirming the alarm\n   state is OK or ALARM (not INSUFFICIENT_DATA) AND the FIS role has\n   `cloudwatch:DescribeAlarms` on the ARN.** Both checks are required;\n   either missing means the stop condition is silently inactive. The\n   checklist MUST cite the alarm state and the role's alarm permission.\n\n4. **NEVER mark IAM role `[✓]` without citing that the action\n   permissions carry tag conditions** (e.g.,\n   `aws:ResourceTag\u002Ffis-target=true`). Without the condition, the role\n   is the only line of defense — and over-scoped roles are the second-\n   most-common FIS incident.\n\n5. **NEVER mark budgetDuration `[✓]` without a one-line rationale\n   tied to the action.** A 30-minute budget on a stop-instances drill\n   is 30 minutes of stopped production; a 30-second budget on a\n   network blackhole may be too short for the SSM command to land.\n   The checklist MUST include the rationale.\n\n6. **NEVER emit `VERDICT: PREREQUISITES_MISSING` without citing the\n   specific gap.** Each `[✗]` item MUST have a one-line reason:\n   `[✗] Stop condition: alarm fis-stop-rate not found — create alarm\n   before applying`. A bare `[✗]` is non-compliant.\n\n7. **NEVER include `aws:ec2:terminate-instances` in an action list\n   without a `[WARN]` note in the checklist.** Terminate is\n   irreversible. The checklist MUST cite that stateful resources will\n   be lost.\n\n### Perfect example output — READY_TO_DEPLOY\n\n```text\nEXPERIMENT_TEMPLATE: ec2-stop-canary (action: aws:ec2:stop-instances, target: fis-target=true)\nVERDICT: READY_TO_DEPLOY\nCHECKLIST:\n  [✓] Intent + blast radius: Stop one canary instance for 60s to validate auto-recovery, blast radius 1 instance\n  [✓] Target scoped: resourceTags fis-target=true (1 instance i-0abc123)\n  [✓] Action(s): aws:ec2:stop-instances (duration PT60S)\n  [✓] Stop condition(s): fis-stop-error-rate (state: OK)\n  [✓] IAM role: fis-execution-role (fis trust, ec2:StopInstances + StartInstances scoped to aws:ResourceTag\u002Ffis-target=true, cloudwatch:DescribeAlarms on fis-stop-error-rate)\n  [✓] Logging: S3 fis-logs-111111111111 + CloudWatch Logs \u002Faws\u002Ffis\u002Fec2-stop-canary\n  [✓] budgetDuration: PT2M (rationale: 2x the 60s stop action for rollback margin)\nVERIFICATION_COMMANDS:\n  aws fis get-experiment-template --id EXT-AAAA1111\n  aws ec2 describe-instances --filters Name=tag:fis-target,Values=true --query 'Reservations[0].Instances[0].InstanceId'\n  aws cloudwatch describe-alarms --alarm-names fis-stop-error-rate --query 'MetricAlarms[0].StateValue'\n  aws iam get-role-policy --role-name fis-execution-role --policy-name fis-execution-policy\n  aws logs describe-log-groups --log-group-name-prefix \u002Faws\u002Ffis\u002Fec2-stop-canary\n```\n\n### Perfect example output — PREREQUISITES_MISSING\n\nFull PREREQUISITES_MISSING example output moved to references.\n→ [references\u002Fworked-examples.md](references\u002Fworked-examples.md) § Perfect example output — PREREQUISITES_MISSING\n\n## Recent AWS features\n\n2024-2026 FIS feature notes (Aurora failover, EKS actions, SSM-based network disruption, org-ID conditions) moved to references.\n→ [references\u002Fadvanced-patterns.md](references\u002Fadvanced-patterns.md) § Recent AWS features\n\n## AWS documentation\n\n- **AWS Fault Injection Service User Guide** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002F\n- **FIS Actions Reference** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002Ffis-actions-reference.html\n- **FIS Target Types** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002Ftargets.html\n- **FIS Stop Conditions** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002Fstop-conditions.html\n- **FIS IAM Roles** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002Fiam-service-role.html\n- **FIS Logging** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002Fuserguide\u002Flogging.html\n- **FIS API Reference** — https:\u002F\u002Fdocs.aws.amazon.com\u002Ffis\u002Flatest\u002FAPIReference\u002F\n- **FIS CLI Reference** — https:\u002F\u002Fdocs.aws.amazon.com\u002Fcli\u002Flatest\u002Freference\u002Ffis\u002F\n- **Blog: AWS Resilience Hub + FIS** — https:\u002F\u002Faws.amazon.com\u002Fblogs\u002Fmt\u002Ftest-application-resilience-using-aws-resilience-hub-and-aws-fault-injection-service\u002F\n- **Workshop: chaos engineering with FIS** — https:\u002F\u002Fcatalog.workshops.aws\u002Fchaosengineering\n\n## References (load on demand)\n\nConsult these only when the corresponding topic comes up:\n\n- [references\u002Ffault-action-catalog.md](references\u002Ffault-action-catalog.md) — fault action IDs, parameters, and rollback semantics\n- [references\u002Fiam-and-logging-templates.md](references\u002Fiam-and-logging-templates.md) — execution-role policies (moved from Step 5) and logging config (moved from Step 6)\n- [references\u002Fworked-examples.md](references\u002Fworked-examples.md) — the PREREQUISITES_MISSING full-output example (moved from § Output format)\n- [references\u002Fadvanced-patterns.md](references\u002Fadvanced-patterns.md) — expert heuristics (permissive-target myth, stop-condition no-op, budgetDuration) and recent AWS features\n",{"name":5,"license":15,"metadata":24,"description":7,"compatibility":43},{"tags":25,"phase":26,"author":27,"domain":28,"family":29,"version":30,"keywords":31,"task_type":32,"complexity":33,"entry_point":34,"skill_class":35,"when_to_use":36,"dependencies":37,"requires_llm":38,"verdict_shape":39,"lifecycle_status":40,"invocation_schema":41,"supports_pipeline":38,"activation_triggers":42},"aws, fis, fault-injection, chaos-engineering, experiment-template, stop-condition, aurora-failover, eks-disruption, deploy, management","1","Jacky Chan — AWS Community Builder","aws-cloudops","Management","0.1.0","aws, fis, fault injection, chaos engineering, experiment template, stop condition, alarm-based stop, aurora failover, rds failover, eks pod disruption, ecs stop-task, ec2 stop-instances, ec2 send-api-error, network blackhole, network latency, network loss, lambda invoke-async, game day, resilience validation, cloudops, deploy","deploy","high","false","capability","Provisioning an FIS experiment template for a game day, DR drill, or resilience gate; configuring CloudWatch alarms as stop conditions; scoping IAM permissions for FIS to run a specific action against a specific target; setting up Aurora failover, EKS pod disruption, ECS task stop, EC2 API error injection, network blackhole\u002Flatency\u002Floss, or Lambda async invocation experiments; or wiring experiment logging to S3 and CloudWatch Logs. Do NOT invoke for production traffic reversal, incident command, or post-incident RCA — use the appropriate troubleshoot or operate skill.","aws-orchestrator","true","READY_TO_DEPLOY | PREREQUISITES_MISSING","active","Input: either (a) an experiment spec including target resource tags\u002F ARNs\u002Ffilters, one or more fault actions with parameters, optional CloudWatch alarm stop conditions, optional logging config, and budgetDuration; or (b) a partial spec for interactive refinement (e.g., \"FIS experiment to fail over an Aurora cluster\"). Output: a deterministic EXPERIMENT_TEMPLATE \u002F VERDICT \u002F CHECKLIST \u002F VERIFICATION_COMMANDS block per the STRICT output contract, where VERDICT is READY_TO_DEPLOY or PREREQUISITES_MISSING.","FIS experiment template, fault injection experiment, chaos engineering, game day drill, resilience validation, Aurora failover test, RDS failover experiment, EKS pod disruption, ECS stop task experiment, EC2 stop instances experiment, EC2 send API error, network blackhole experiment, network latency injection, network packet loss, Lambda invoke async experiment, FIS stop condition, alarm-based stop","Agent runtime that reads SKILL.md (Claude Code, Cursor, Windsurf, Codex, Gemini). Live provisioning uses AWS CLI v2 with fis (create-experiment-template, get-experiment-template, start-experiment), iam (create-role, put-role-policy), cloudwatch (describe-alarms), s3 (create-bucket, put-bucket-policy) and logs (create-log-group). Works with Terraform aws_fis_experiment_template \u002F aws_iam_role and CloudFormation AWS::FIS::ExperimentTemplate.","main","eeace51bd5dd1db2030729d83eeab988861553aa","skills\u002Ffis-experiment-deployer",0,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills","https:\u002F\u002Fgithub.com\u002Fjhkchan","CTO & Co-Founder @Beever-AI  & @votee. CAISP. AWS Community Builder. Organizer@GDGCloudHK. MongoDB Champion. Technopreneur + Infornographer.","2026-09-24 08:54:10",{"source":8,"skillName":5,"commands":53,"prompt":63,"zipAvailable":64},[54,57,60],{"tool":55,"command":56},"npx","npx skills add jhkchan\u002Faws-agent-skills --skill fis-experiment-deployer",{"tool":58,"command":59},"bunx","bunx skills add jhkchan\u002Faws-agent-skills --skill fis-experiment-deployer",{"tool":61,"command":62},"pnpm","pnpm dlx skills add jhkchan\u002Faws-agent-skills --skill fis-experiment-deployer","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002FSKILL.md",true,[66,72,76,80,84,88,92,96,100,104,108,112,116,120,124,128,132,136,140],{"path":67,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":70,"sourceHtmlUrl":71},"eval\u002Ftest-cases.yaml","REFERENCE","参考文档",5679,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Feval\u002Ftest-cases.yaml",{"path":73,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":74,"sourceHtmlUrl":75},"evals\u002Fbaselines\u002Faurora-failover-dr-drill.md",1052,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Fbaselines\u002Faurora-failover-dr-drill.md",{"path":77,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":78,"sourceHtmlUrl":79},"evals\u002Fbaselines\u002Fec2-stop-canary-with-alarm.md",1010,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Fbaselines\u002Fec2-stop-canary-with-alarm.md",{"path":81,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":82,"sourceHtmlUrl":83},"evals\u002Fbaselines\u002Feks-pod-disruption.md",1261,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Fbaselines\u002Feks-pod-disruption.md",{"path":85,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":86,"sourceHtmlUrl":87},"evals\u002Fbaselines\u002Fnetwork-blackhole-with-ssm.md",1214,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Fbaselines\u002Fnetwork-blackhole-with-ssm.md",{"path":89,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":90,"sourceHtmlUrl":91},"evals\u002Fbaselines\u002Fpermissive-target-tag-rejected.md",874,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Fbaselines\u002Fpermissive-target-tag-rejected.md",{"path":93,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":94,"sourceHtmlUrl":95},"evals\u002Feval_config.yaml",1787,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Feval_config.yaml",{"path":97,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":98,"sourceHtmlUrl":99},"evals\u002Fevals.json",13242,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Fevals.json",{"path":101,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":102,"sourceHtmlUrl":103},"evals\u002Fprompts\u002Faurora-failover-dr-drill.md",549,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Fprompts\u002Faurora-failover-dr-drill.md",{"path":105,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":106,"sourceHtmlUrl":107},"evals\u002Fprompts\u002Fec2-stop-canary-with-alarm.md",627,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Fprompts\u002Fec2-stop-canary-with-alarm.md",{"path":109,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":110,"sourceHtmlUrl":111},"evals\u002Fprompts\u002Feks-pod-disruption.md",523,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Fprompts\u002Feks-pod-disruption.md",{"path":113,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":114,"sourceHtmlUrl":115},"evals\u002Fprompts\u002Fnetwork-blackhole-with-ssm.md",610,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Fprompts\u002Fnetwork-blackhole-with-ssm.md",{"path":117,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":118,"sourceHtmlUrl":119},"evals\u002Fprompts\u002Fpermissive-target-tag-rejected.md",365,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fevals\u002Fprompts\u002Fpermissive-target-tag-rejected.md",{"path":121,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":122,"sourceHtmlUrl":123},"example\u002FREADME.md",7603,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fexample\u002FREADME.md",{"path":125,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":126,"sourceHtmlUrl":127},"references\u002Fadvanced-patterns.md",5430,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Freferences\u002Fadvanced-patterns.md",{"path":129,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":130,"sourceHtmlUrl":131},"references\u002Ffault-action-catalog.md",7240,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Freferences\u002Ffault-action-catalog.md",{"path":133,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":134,"sourceHtmlUrl":135},"references\u002Fiam-and-logging-templates.md",9922,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Freferences\u002Fiam-and-logging-templates.md",{"path":137,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":138,"sourceHtmlUrl":139},"references\u002Fworked-examples.md",1669,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Freferences\u002Fworked-examples.md",{"path":141,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":142,"sourceHtmlUrl":143},"skill.usf.yaml",4507,"https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffis-experiment-deployer\u002Fskill.usf.yaml",[145,152,159,166,173,180],{"skillNo":146,"name":147,"nameZh":6,"description":148,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":149,"categories":150,"keywords":151,"publishedAt":19},"01M7ADVD60X4PCAPKSE58P9NDC","xray-tracing-deployer","Deploys AWS X-Ray distributed tracing with production-grade configuration: X-Ray daemon deployment (EC2 systemd, ECS sidecar, EKS DaemonSet, Lambda built-in), SDK instrumentation (Java, Python, Node.js, Go, .NET) with annotations and metadata, sampling rules (default reservoir + rate vs custom rules with URL \u002F method \u002F service-name predicates), service map generation, trace query and retrieval, groups and insights, CloudWatch ServiceLens integration, IAM permissions (PutTraceSegments, GetSamplingRules), and latest features (Lambda Powertools tracing, OpenTelemetry \u002F ADOT Collector support, W3C trace context). Emits a READY_TO_DEPLOY checklist with every configuration item verified. Use when enabling X-Ray on a new application, adding tracing to ECS \u002F EKS \u002F Lambda, validating a sampling configuration, or generating daemon deployment CLI and IaC templates. Triggers: X-Ray, distributed tracing, X-Ray daemon, sampling rules, service map, X-Ray SDK, OpenTelemetry, ADOT, Lambda Powertools, CloudWatch ServiceLens.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fxray-tracing-deployer\u002FSKILL.md",[],[],{"skillNo":153,"name":154,"nameZh":6,"description":155,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":156,"categories":157,"keywords":158,"publishedAt":19},"01M7ADTVJ7DCBM9CB1FY51GH7J","wellarchitected-workload-auditor","Audits AWS Well-Architected Tool workloads for review staleness, per-pillar high-risk issue counts, milestone tracking gaps, and remediation plan completeness. Emits a deterministic verdict (STALE_REVIEW | HIGH_RISK | CONFIG_GAP | OK) per workload with per-pillar risk breakdown, milestone trend analysis, and specific remediation. Use when reviewing Well-Architected workloads, checking WA review freshness, auditing high-risk issues per pillar, validating milestone coverage, or tracking remediation plans.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwellarchitected-workload-auditor\u002FSKILL.md",[],[],{"skillNo":160,"name":161,"nameZh":6,"description":162,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":163,"categories":164,"keywords":165,"publishedAt":19},"01M7ADT59G6FTTNZYHT2QRPNNP","wellarchitected-review-operator","Operates AWS Well-Architected Tool reviews end-to-end — creates workloads, runs pillar reviews (operational excellence, security, reliability, performance, cost optimization, sustainability, Prosperity), answers questions with risk-tier rationale, generates reports and improvement plans, creates milestones, and integrates with Trusted Advisor and Well-Architected Labs. Runs deterministic pre-checks (workload existence, lens availability, pillar coverage), emits the exact update-answer \u002F get-consolidated-report CLI behind a CONFIRM gate, and verifies risk-tier transitions. Emits READY | BLOCKED | COMPLETED. Use when creating a workload, answering pillar questions, generating an improvement plan, creating a milestone, or integrating with Trusted Advisor. Triggers: well-architected tool, well architected review, workload review, pillar, operational excellence, security, reliability, performance, cost optimization, sustainability, prosperity, improvement plan, milestone, trusted advisor.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwellarchitected-review-operator\u002FSKILL.md",[],[],{"skillNo":167,"name":168,"nameZh":6,"description":169,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":170,"categories":171,"keywords":172,"publishedAt":19},"01M7ADSH5JA9QHNPPHQCD7EVA7","wafv2-web-acl-deployer","Provisions AWS WAFv2 Web ACLs with correct production defaults: scope selection (CLOUDFRONT vs REGIONAL), managed rule groups (AWSManagedRulesCommonRuleSet, SQLiRuleSet, LinuxRuleSet, WindowsRuleSet, IPReputationList, BotControlRuleSet, AmazonIpReputationList, ATP), custom rules (byte match, IP set, regex pattern, geo match, size constraint, rate-based), rule priority ordering, logging destinations (Kinesis Firehose, CloudWatch Logs, S3), resource association (ALB, API Gateway, CloudFront), rate-based rules with aggregate keys (IP, forwarded IP), Challenge and CAPTCHA actions, and account takeover prevention. Emits a READY_TO_DEPLOY checklist. Use when creating a Web ACL, attaching managed rules, configuring rate limiting, enabling WAF logging, associating a Web ACL with an ALB\u002FAPI Gateway\u002FCloudFront, or configuring CAPTCHA\u002FChallenge actions. Triggers: create WAF Web ACL, WAFv2, managed rule groups, WAF rate limiting, WAF logging, WAF CAPTCHA, WAF Challenge action, bot control, ATP managed rules.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwafv2-web-acl-deployer\u002FSKILL.md",[],[],{"skillNo":174,"name":175,"nameZh":6,"description":176,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":177,"categories":178,"keywords":179,"publishedAt":19},"01M7ADRYFAT8R73FBNCHRC573X","wafv2-web-acl-auditor","Audits AWS WAFv2 Web ACL configurations to determine whether the ACL provides effective protection: default-action posture (Allow vs Block), managed-rule-group coverage gaps, rule effectiveness (BLOCK vs COUNT, shadow\u002Fbypass rules, stale exclusions), rate-based rule correctness, logging and visibility configuration, and text-transformation bypass vectors. Emits a deterministic verdict per Web ACL. Use when reviewing a WAFv2 Web ACL for security posture, checking managed-rule coverage, validating rate-limit configuration, auditing WAF logging, or hardening a WAF before production deployment. Triggers: WAF, WAFv2, Web ACL, managed rule group, AWSManagedRules, rate-based rule, CAPTCHA, Challenge, OverrideAction, Count mode, forwarded IP, Bot Control, ATP, text transformation, WAF logging, WAF visibility, OWASP, SQLi, XSS, RFI.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwafv2-web-acl-auditor\u002FSKILL.md",[],[],{"skillNo":181,"name":182,"nameZh":6,"description":183,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":184,"categories":185,"keywords":186,"publishedAt":19},"01M7ADR8W7G72JX4KFGN03QK8Q","waf-rule-deployer","Provisions AWS WAFv2 rule sets with production defaults: Web ACL creation (CloudFront scope in us-east-1 vs Regional scope), rule groups (managed vs custom) with capacity planning in WCU (Web ACL Capacity Units, 1500 default per ACL), managed rule groups (AWSManagedRulesCommonRuleSet, AWSManagedRulesSQLiRuleSet, AWSManagedRulesAmazonIpReputationList, AWSManagedRulesLinuxRuleSet, AWSManagedRulesWindowsRuleSet, AWSManagedRulesUnixRuleSet), custom rules (byte-match, regex pattern set, geo-match, rate-based), IP set and regex pattern set resources, rule priority ordering (lowest number evaluated first), action types (allow, block, count, CAPTCHA, Challenge), bot control managed rule group, account takeover. Triggers: create WAF rule, WAF managed rule group, WAF custom rule, WAF byte match, WAF geo match, WAF rate based rule, WAF IP set, WAF regex pattern set, WAF WCU budget, WAF CloudFront ACL, WAF logging Kinesis Firehose, WAF bot control, WAF ATP, WAF label matching, WAF Challenge action, WAF CAPTCHA.","https:\u002F\u002Fgithub.com\u002Fjhkchan\u002Faws-agent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwaf-rule-deployer\u002FSKILL.md",[],[]]