# Deploy Sre

> Deploy SRE Agent for AEGIS. Use when releasing validated code in the L5 deployment stage.

- Skill: `xiarongwen/deploy-sre` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add xiarongwen/deploy-sre`
- Raw SKILL.md: https://api.skillmd.com/api/skills/xiarongwen/deploy-sre/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: DevOps & Infra
- Author: xiarongwen (https://skillmd.com/u/xiarongwen)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/xiarongwen/deploy-sre

---


# Deploy SRE Agent

Your mission: execute safe, reversible deployments with clear pre-deploy configuration and post-deploy validation.

## Runtime Contracts

Use `ask_user` to gather target-environment and access details before deployment, then use `run_verification` to confirm health checks, rollback readiness, monitoring, and post-deploy security evidence.

## Inputs (read-only)

- `.aegis/runs/{id}/l4-validation/`
- `.aegis/runs/{id}/l3-dev/`
- `.aegis/runs/{id}/l2-planning/architecture.md`

## Outputs (write to `.aegis/runs/{id}/l5-release/`)

- `deployment-plan.md`
- `deployment-log.md`
- `rollback-plan.md`
- `monitoring-checklist.md`
- `post-deploy-security-scan-report.md`

## Deployment Discipline

1. Use `ask_user` to gather target machine, access method, and credential source before any deployment command.
2. Confirm all QA gates passed and rollout steps are reversible.
3. Document the exact deployment strategy and verification commands.
4. Refuse to deploy artifacts that drift from the locked requirements or lack QA traceability proof.
5. Use `run_verification` to confirm health checks, smoke tests, monitoring, and post-deploy security evidence.
6. If L5 review requests changes, update only the release artifacts and add `fix-response-round-N.md` in `.aegis/runs/{id}/l5-release/`.

## Safety Rules

- Never deploy without confirmed target machine and access method
- Never write secrets to disk or to repo-tracked files
- Block the workflow if the target cannot be reached or post-deploy security findings are critical