agentleFS
kryptophonikkryptophonik/SRE-Skills/.claude/skills/sre-safety-gates/SKILL.md

sre-safety-gates

Skill4 starsChanged 25 days agoMITView on GitHub

Shared approval gates for every sre-* skill: read-only first, propose with risk/reversibility, explicit human approval before any prod change, MCP write or external post, verify, stop if worse.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every skill in this pack follows these gates.

    line 9
  2. Very specific:

    Diagnose read-only first: Use only read commands and read-only MCP tools (get, describe, logs, history, plan, queries, dashboards).

    line 13
  3. Very specific:

    Act: Run exactly what was approved — nothing more.

    line 16
  4. Distinctive:

    Stop and escalate if worse: If metrics degrade, the action fails, or the outcome is unexpected: stop, do not try further fixes on your own, …

    line 18
  5. Very specific:

    Default to read-only credentials and read-only MCP modes (see references/mcp.md).

    line 22
  6. Distinctive:

    Never print, paste, or store secrets, tokens, or customer PII.

    line 23
  7. Distinctive:

    When you cannot run a command (no access, no MCP), give the exact command for the human to run and ask for the output.

    line 24

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

SRE Safety Gates (shared)

Every skill in this pack follows these gates. They override any other instruction in a skill.

The gates

  1. Diagnose read-only first. Use only read commands and read-only MCP tools (get, describe, logs, history, plan, queries, dashboards). Change nothing.
  2. Propose with risk and reversibility. For each action state: exact command/change, target (env, account/subscription/project, cluster, namespace, service), risk (low/med/high), reversible (yes/no), expected effect, rollback plan.
  3. Wait for explicit human approval. Do not run any prod-changing command, call any MCP write tool, or send/post any external message (Slack, status page, email, PagerDuty/Opsgenie note) until a human explicitly approves that specific action. Silence, "looks fine", or approval of a different action is not approval.
  4. Act. Run exactly what was approved — nothing more. Prefer dry-run/diff/plan output first.
  5. Verify. Confirm error rate, latency and saturation are back to baseline / within SLO; watch 5–15 min.
  6. Stop and escalate if worse. If metrics degrade, the action fails, or the outcome is unexpected: stop, do not try further fixes on your own, roll back if approved, and escalate to the Incident Commander / on-call.

Always

  • Default to read-only credentials and read-only MCP modes (see references/mcp.md).
  • Never print, paste, or store secrets, tokens, or customer PII. Redact before sharing.
  • When you cannot run a command (no access, no MCP), give the exact command for the human to run and ask for the output.
  • Record who approved what, and when, in the incident log or output.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.