More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Async Workflow Completion Repair

    1

    It performs the supplied case directly, separates verified evidence from unknowns, produces bounded actions, and ties closure to case specific proof rather than generic advice.

    $1

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Async Workflow Completion Repair

    Async Workflow Completion Repair

    Example session with this skill installed

    A document export job sets status to Complete after writing the main file, while thumbnail generation and notification delivery continue asynchronously. If either later fails, the UI still shows Complete and retries can run after the customer has downloaded the package.

    • Read your context and instructions
    • Compiled the async workflow completion

    ASYNC WORKFLOW COMPLETION REPAIR

    Case analyzed

    A document export job sets status to Complete after writing the main file, while thumbnail generation and notification delivery continue asynchronously. If either later fails, the UI still shows Complete and retries can run after the customer has downloaded the package.

    This review uses only the supplied facts. It does not assume access to production systems, private records, customer systems, vendor portals, or approvals that were not provided.

    Verdict

    NOT READY TO CLOSE. The case contains a material control or evidence gap that must be resolved before the workflow can be treated as complete.

    Verified evidence

    • VERIFIED: The scenario and constraints above were explicitly supplied.
    • VERIFIED: At least one required ownership, sequencing, evidence, or validation condition is missing or inconsistent.
    • VERIFIED: No additional approval or successful execution was supplied.
    • UNKNOWN: Any fact not present in the case remains unknown and must not be invented.

    Concrete findings

    1. The supplied case contains enough evidence to identify a material control or readiness gap, but not enough to claim the workflow is safe or complete.
    2. The highest risk is the mismatch between the stated operating intent and the evidence or state described in the case.
    3. Any remediation must preserve the historical record and keep verified facts separate from assumptions introduced during repair.
    4. Closure must depend on deterministic evidence tied to this exact case rather than a status field, optimistic label, or verbal confirmation.

    The controlling rule for this case is

    An asynchronous workflow is complete only when every required side effect has a durable terminal outcome, retries cannot reopen closed state, and completion is reconstructable from evidence.

    Required actions

    1. Capture the current state for completion state machine and identify the accountable owner.
    2. Reconcile required side effects against the intended rule and document every material mismatch.
    3. Create a bounded correction for retry and callback ordering without rewriting or inventing historical evidence.
    4. Validate terminal failure handling using a repeatable check tied to the supplied scenario.
    5. Do not close the case until durable completion evidence has explicit evidence and any remaining exception is named.

    Deterministic validation

    • A reviewer can reproduce the conclusion from the supplied evidence without relying on private context.
    • Every material mismatch has an owner and a specific corrective action.
    • Unknown facts remain labeled UNKNOWN rather than being inferred as completed work.
    • The final state can be checked independently against the stated acceptance condition.
    • No production verification, approval, customer acceptance, or execution result is claimed unless it was actually supplied.

    Evidence boundaries

    • UNKNOWN: Details not explicitly present in the supplied case.
    • UNKNOWN: Whether proposed corrections have been executed.
    • UNKNOWN: Whether downstream owners have accepted the repaired state.
    • UNKNOWN: Whether production behavior matches the intended control.

    These unknowns define the exact evidence required before closure without guessing.

    Decision and handoff

    Keep the case open until the required evidence is attached, the responsible owner is identifiable, and the acceptance condition is demonstrably satisfied. Preserve the original record and append corrections rather than silently rewriting history.

    If the workflow changes later, retain the superseded reasoning and link the new decision to the evidence that caused the change. A later reviewer should be able to reconstruct the chain from request to evidence, decision, action, and verification.

    Closure rule

    Close only when the requested result is supported by evidence. A completed status, saved form, successful navigation, or verbal assurance is not sufficient on its own. The closure record must show what was checked, what changed, who owns the result, and which condition was actually satisfied.

    Acceptance detail

    The final review should explicitly revisit these dimensions: completion state machine, required side effects, retry and callback ordering, terminal failure handling, durable completion evidence. For each one, record the source evidence, current state, expected state, difference, owner, and validation result. If any dimension remains unknown, keep it open or record a bounded exception with an expiry and owner.

    Final conclusion

    The correct disposition for this supplied case is to continue remediation, not to declare success. The evidence is sufficient to identify the gap and define the next actions, but it is not sufficient to claim the repaired state already exists.

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

    About this skill

    Async Workflow Completion Repair is a ToolForge Labs workflow to repair asynchronous workflows that mark work complete before all required side effects, callbacks, retries, or terminal outcomes are durably known. It performs the supplied case directly, separates verified evidence from unknowns, produces bounded actions, and ties closure to case specific proof rather than generic advice.

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

    Cannot verify external systems or facts not present in supplied evidence. With partial context it returns explicit evidence gaps instead of fabricated outcomes.

    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 month ago

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    Use a controlled workspace or supplied evidence. Review owner decisions and proposed changes before operational action.

    Listed1 month ago
    Updated9 days ago

    What's inside

    Frequently Asked Questions

    This skill prevents 'silent failures' by ensuring a workflow only marks itself as complete once all side effects, like database writes and third-party webhooks, are verified as durable and terminal.

    The skill is framework-agnostic and supports any architecture using task queues, event loops, or orchestrators like Temporal, AWS Step Functions, and RabbitMQ.

    You will receive a specialized AI agent configuration designed to audit state machines, identify race conditions, and generate specific corrective logic for distributed workflows.

    No, the skill operates through static analysis of logs, code snippets, and architecture diagrams you provide, ensuring no unauthorized changes are made to your live infrastructure.

    It solves for race conditions where background retries fire after a process is closed, or where a 'Success' status is returned while a critical sub-task is still pending or has quietly failed.

    New in Agents & Orchestration