skillshop-ostyles/agents-operations

Backup coverage scanner: inventory every stateful resource, trace backup configuration for each, then LLM identifies critical gaps. Read-only. Trigger: /backup-scan

Capacity early warning: find hardcoded limits, pool sizes, timeouts, quotas, then LLM judges each as adequate/approaching/critical. Read-only. Trigger: /capacity

CI debt analyzer: read CI configuration (GitHub Actions, GitLab CI, Jenkins, CircleCI), measure pipeline health, then LLM judges what is costing the team most. Read-only. Trigger: /ci-debt

Dependency graveyard: inventory every dependency, check registry health metadata, then LLM judges each as healthy/aging/zombie/dead. Read-only. Trigger: /dep-graveyard

Deployment drift detector: extracted deployed config (kubectl, terraform show, docker inspect) vs source-of-truth manifests, then LLM judges each drift's criticality in business context. Read-only. Trigger: /deploy-drift

Deployment frequency tracker: compute DORA metrics from git history, LLM identifies bottlenecks and improvement opportunities. Read-only. Trigger: /deploy-freq

Dockerfile best-practices auditor: statically scans Dockerfiles for 18 common anti-patterns including unpinned base images, root execution, missing HEALTHCHECK, excessive layers, package cache bloat, and hardcoded secrets. Produces an evidence-backed report with severity and remediation. Read-only. Trigger: /dockerfile-audit

Env drift tracker: compare config values across environments (dev/staging/prod), LLM flags each difference with risk assessment. Read-only. Trigger: /env-drift

Failure simulator on code level: inventories every external touchpoint (HTTP clients, DB access, filesystem, queues, caches) with its surrounding error handling, then for a chosen failure scenario (DB down, API timeouts, disk full) mentally executes the failure path at each touchpoint and reports the resulting behavior - retry, degradation, crash or silent loss - plus inconsistencies and hardening recommendations. Pure thought experiment, nothing is ever shut down. Read-only. Trigger: /failsim

Healthcheck mapper: inventory all health/readiness/liveness endpoints, map against service dependencies, LLM judges each as adequate/weak/missing. Read-only. Trigger: /healthcheck

Leak detector: trace resource acquisition and release across code paths, LLM classifies each as clean/leaky/uncertain. Read-only. Trigger: /leak-scan

Log quality auditor: inventory every log statement, check for structure, correlation IDs, levels, PII risk, then LLM judges operational quality. Read-only. Trigger: /log-audit

Rollback readiness: check each deployable change against rollback criteria, LLM estimates cost and risk of undoing it. Read-only. Trigger: /rollback

Runbook auditor: read runbook files, extract verifiable claims, check each against current codebase. LLM judges correctness and completeness. Read-only. Trigger: /runbook-audit

Time bomb scanner: finds hardcoded dates, expiry deadlines, cert references, 32-bit time usage and 'temporary' markers rotting since years (git age via blame), then has the LLM classify each finding as live bomb / rotten provisional / false alarm and produce a defusal list ranked by detonation date. Read-only. Trigger: /timebomb