escalation ladder

    1

    A runtime risk-scoring engine that routes agent actions to ACT, ASK, or STOP rungs based on safety dimensions.

    $9

    /mo

    Subscription · cancel anytime

    Secure checkout via Stripe

    • Always the latest version
    • Delivered live to your agent via MCP
    • Cancel anytime, access ends at period end
    Security scanned

    Works with the AI tools you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+22 more

    escalation ladder

    Example session with this skill installed

    I'm about to delete 50 stale branches in the production repo and then email the team leads about the cleanup. Score this and provide the rung.

    • Read your context and instructions
    • Compiled the escalation ladder

    I have evaluated your proposed plan across two distinct actions. The overall plan has been escalated to

    STOP due to the high blast radius of production branch deletion, though individual components vary.

    Escalation Assessment

    ActionReversibilityMoneyDataPeopleBlast RadiusAmbiguityRung
    Delete 50 production branches413352STOP
    Email team leads312321ASK

    Decision Logic

    Delete 50 production branches: This action is scored as a

    5 (STOP) for Blast Radius and a

    4 (STOP) for Reversibility. Deleting production resources is a high-stakes, irreversible action that could impact CI/CD pipelines and deployment history.
    2.

    Email team leads: This scores a

    3 (ASK) for Reversibility and People. Once sent, the communication cannot be retracted from the recipients' awareness, potentially affecting team workflow.

    Escalation Response

    I will not proceed with the deletion of production branches. This action violates safety protocols for automated production modifications.

    Proposed Action: Send a summary email to team leads regarding suggested branch cleanups.
    My Understanding: You want to reduce repository clutter to improve developer experience.
    Options:

    1. Draft only: I will generate the email draft and the list of 50 branches for your manual review.

    Scheduled Notice: I will email the leads a list of branches marked for deletion in 7 days, allowing them to opt-out.
    Recommendation: Option 1. Given the production context, a manual review of the branch list is the safest path.
    Default: No action will be taken if you do not respond.

    Next steps

    • Review the list of 50 branches I have identified as "stale."
    • Confirm if you would like me to proceed with Option 1 (Drafting the notice).
    • Provide an explicit override if you require immediate production deletion.
    $9/mo

    Subscription · always the latest version

    • Secure checkout via Stripe
    • Cancel anytime
    • Delivered live via MCP (optional)

    What you get

    Prevent unauthorized API spending or high-cost model calls.Audit and gate irreversible file system or database deletions.Standardize human-in-the-loop checkpoints for customer-facing communications.Enforce strict boundaries on sensitive data exfiltration and credential access.

    About this skill

    The problem

    Agents often fail by acting when they should have asked, or asking when they should have stopped. A lack of a runtime escalation policy leads to unauthorized purchases, data deletion, and irreversible communication errors.

    What it does

    • Evaluates every proposed agent action against six risk dimensions: reversibility, money, data, people, blast radius, and ambiguity.
    • Routes actions to one of three rungs: ACT (execute), ASK (human-in-the-loop decision), or STOP (outright refusal).
    • Standardizes the human-in-the-loop experience with a five-part ASK format including options, recommendations, and default behaviors.
    • Prevents "permission theater" for low-stakes tasks while hard-blocking high-risk maneuvers like credential exfiltration or self-modification.

    Why this beats prompting it yourself

    Vague instructions like "ask for permission if unsure" lead to inconsistent agent behavior and decision fatigue. This skill replaces "vibes" with a concrete scoring engine that treats uncertainty as a trigger for escalation rather than an excuse for recklessness.

    Use cases

    • Financial agents managing budgets who need a hard floor for unverified recipients.
    • DevOps agents performing batch operations on production-adjacent infrastructure.
    • Customer support agents drafting messages that require sign-off based on brand reputation risk.
    • Personal assistants managing sensitive data across different trust boundaries.

    Known limitations

    Requires an agent capable of pausing execution for human feedback. Does not provide a UI for the human response, only the logic for the request.

    How to install

    Works the same in every agent - Claude, Cursor, Codex, Copilot and 20+ more.

    ~30 seconds
    1. 1

      Download the ZIP

      Free skills download straight away. Paid skills unlock right after purchase.

    2. 2

      Unzip into your skills folder

      Every agent reads skills from one folder on your machine. Drop the unzipped folder in there.

    3. 3

      Ask your agent to use it

      Restart the agent if it was already running. It picks the skill up automatically - no config needed.

    Skills folder by agent

    Click the path to copy it. Create the folder if it does not exist yet.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill passes our trust checks below.

    Security scanned

    Passed our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scanned

    Verified clean 14 days ago

    Listed14 days ago

    What's inside

    Frequently Asked Questions