Pre-Delivery AI Self-Check
Pre-Delivery AI Self-Check
Showcase
Description
Perform AI pre-delivery self-checks on various tasks, proactively fix resolvable issues, and honestly explain unfinished items and genuine blockers.
Best for
Individuals or teams needing to verify completeness, results, and limitations before delivery.
How to use this Skill
Provide your input
Provide the current task requirements, conversation background, completed work, available tools, and the proposed full response.
Run the Skill
Review the task scope, completion status, evidence, permissions, and verification before delivery; safely resolve and re-check fixable issues, and explain the reasons, impact, and necessary next steps for unfinished items.
Review your result
Returns a reviewed result matching the original task, noting unfinished items, genuine blockers, and required authorizations when necessary, without fabricating completion status or verification results.
Related Skills
View all
Pre-Delivery QC · Bilingual
Perform quality control before a formal document is delivered, submitted, or published, in either Chinese or English. The skill follows an Auto-Flow Prompt (AFP) five-layer workflow: the Trigger Layer identifies the task and creates a record; the Planning Layer generates review items and an execution plan from the routing table; the Execution Layer adaptively reviews each item, including factual claims, certainty levels, task omissions, numerical standards, logic and structure, compliance and privacy, practical usability, audience mismatch, proper nouns, timeliness, layout and formatting, figures, links, attachments, collaboration traces, and more. The Supervision Layer runs a “check → challenge → verify → correct → retest” self-review loop and flags items requiring human intervention. The Integration Layer removes duplicates, assigns priority levels, resolves conflicts across items, and consolidates the results into a report. The final output includes a P0/P1/P2 issue list with each item’s location, basis, and suggested action, along with an overall verdict: “Releasable / Releasable after revision / Not releasable.” The skill automatically detects the document’s language and produces the report in that language. It reports only: it does not modify files or automatically send or publish them. Final release authority and responsibility remain with a human.

AI Delivery Check
Check whether AI-generated deliverables (web pages, apps, documents, etc.) fully satisfy all user requirements from the conversation. Review each item across functionality, visuals, and content. If any fall short, trigger AI corrections to ensure the final output aligns with user expectations.
ResearchCloud Service Risk Check
A billing and account risk check for independent developers building products with AI. It covers Cloudflare, Vercel, Supabase, Stripe, VPSs, object storage, and AI APIs such as OpenAI, Gemini, and Anthropic: first verifying whether each platform has any hard limits today that could suspend service, then checking for exposed keys, database permissions, self-scheduling loops, single-account points of failure, and frozen funds. The report only covers what relates to your tech stack: how much the worst case could burn, the cost of applying each brake, the switches to flip today, and prompts you can hand directly to an AI to fix code. If something has already gone wrong, it switches to emergency mode, first telling you what to stop and what to capture, then helping you write an appeal. Principle: count refunds as zero; alerts don't count as brakes.
Information
- Version
- v2
- Last updated
- Runtime credits
- Usage-based
- Models
- Auto