Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Self-Healing Ops Agent: Safe AI Auto-Remediation

    1

    Design an autonomous SRE agent that diagnoses failing jobs, writes a minimal fix in a throwaway sandbox, proves it through quality, security and risk gates, and opens a pull request, never writing to production directly.

    $19

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Self-Healing Ops Agent: Safe AI Auto-Remediation

    Self-Healing Ops Agent: Safe AI Auto-Remediation

    Example session with this skill installed

    Our auto-remediation agent produced the patch in fix.diff for a failing ingestion job and wants to deploy it straight to production because 'tests pass'. Decide what should happen to this change and explain in decision.md.

    • Read your context and instructions
    • Compiled the self-healing ops agent

    decision.md: not approved for autonomous deployment. The diff mixes a legitimate fix (raise UpstreamNotReady when the upstream file is missing) with the removal of the '/ 1000.0' scaling in settlement_pnl, a protected business formula; the change gate marks it HIGH. Split it: the guard goes out as a LOW-risk PR (eligible for a canary only after review and tests), the formula change is dropped or sent to the formula owner for human review. Record both in incident memory.

    decision.md

    Connects securely to your tools. The creator never sees your data.

    About this skill

    The core invariant: the agent never writes to production. It reads evidence, writes a candidate patch in an isolated sandbox, and hands a verified, risk-classified change to your normal delivery pipeline. Autonomy is earned per change class, not granted per agent.

    What your agent builds with it

    • An incident controller in plain code: skips platform faults and deliberate guards, backoff per incident key, attempts recorded before the session, de-duplication.
    • Read-only diagnosis: incident memory first, logs, config, guarded SQL, allow-listed commands; logs and files treated as evidence, never instructions.
    • Fixing in a throwaway clone with hard rules: never remove a deliberate guard, never change protected invariants, never paper over missing upstream data, never weaken a test.
    • A change gate (change_gate.py): per-file risk from path rules, size limits and protected-pattern removals; security scan of added lines; a change-set hash so any later edit voids the approval. Only LOW risk + tests passed + no findings is approved for autonomous rollout.
    • Publishing as PRs on fix/autofix-* branches with a full evidence template; optional canary at 1–5% with automatic rollback, off by default.
    • Honest reporting (RESOLVED only when a real change reached the remote), incident memory, and the metrics that prove it works, including "incidents made worse".

    Included: SKILL.md, architecture reference with the full tool list and rollout plan, an agent-instructions template, an example change policy, change_gate.py (standard library), and an eval case.

    Part of the Production AI Agents series. Pair with Agent Guardrails and Agent Reliability; start free with Agent Architecture.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.

    A design and gating toolkit, not a hosted service. The change gate's security scan is pattern-based and complements, not replaces, your static analysis and dependency scanners. Path and protected-pattern rules must be tuned to your codebase.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 day ago

    Software
    Python 3.10+Git hosting with pull requestsGit
    Optional
    Google Cloud (Cloud Run, Logging, BigQuery)CI/CD with traffic splitting

    Permissions

    File Scopes

    Your
    repository
    (read)
    a
    throwaway
    clone
    for
    patches
    design
    documents
    you
    ask

    change_gate.py reads a diff from a file or stdin and makes no network calls. The skill designs the agent; it does not deploy anything itself.

    Listed1 day ago

    What's inside

    Frequently Asked Questions

    Not by default. It opens pull requests; canary deployment is an opt-in per incident, only for LOW-risk changes whose approval matches the exact diff hash, with automatic rollback.

    Hard rules in the agent instructions plus the change gate: removing a protected pattern or touching blocked or high-risk paths makes the change HIGH or BLOCKED, which always goes to humans.

    New in Security