sre-engineer
PassAudited by ClawScan on May 7, 2026.
Overview
This is a coherent instruction-only SRE skill with no code or credentials, but users should supervise any production-impacting automation, testing, or status claims it helps produce.
This skill is safe to install as an instruction-only SRE assistant, but do not let it directly modify production infrastructure, run chaos/load tests, or report reliability metrics without explicit scope, approvals, and verification.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The skill could help make operational changes that affect service reliability or availability if the user allows it to use real infrastructure tools.
These are purpose-aligned SRE activities, but if an agent has ambient access to infrastructure tools they could change deployments, infrastructure, or automation pipelines. The artifact does not define explicit approval or rollback boundaries.
Implement solutions maximizing reliability while maintaining feature velocity ... Automation development: - Python scripting - Go tool development - Terraform modules - Kubernetes operators - CI/CD pipelines
Use it with scoped credentials and require human review before applying Terraform, Kubernetes, CI/CD, or other production changes.
Improperly scoped reliability tests could create outages, noisy alerts, or load on shared systems.
Chaos, load, and failure testing are expected for an SRE skill and the prompt mentions safety controls, but such tests can still disrupt systems if run without clear scoping.
Chaos engineering: - Experiment design - Hypothesis formation - Blast radius control - Safety mechanisms ... Production readiness: ... - Load testing - Failure testing
Run tests only in approved environments or with explicit blast-radius limits, rollback plans, and monitoring.
A user could be misled by precise-looking reliability metrics if they are reported without evidence.
The prompt includes a specific quantified completion message. It appears to be illustrative, but users should ensure the agent does not present these numbers as verified results unless they were actually measured.
Delivery notification: "SRE implementation completed. Established SLOs for 95% of services, reduced toil from 70% to 35%, achieved 24-minute MTTR, and built 87% automation coverage."
Treat the quoted delivery message as a template only and require citations to measured data for any reported SLO, toil, MTTR, or automation-coverage numbers.