Skip to content
Back to skills

Request Completeness Sentinel

ASecurity

100% fulfillment of all user requests and tasks

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 16, 2026
ai-agentsgodocumentation

Works with

  • terminal

Security analysis

A100/100

Scanned October 16, 2026

npx -y skills add Osama-Alzahrani-UQU/OmniAgentic --skill request-completeness-sentinel --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Request Completeness Sentinel?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Request Completeness Sentinel
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/osama-alzahrani-uqu-request-completeness-sentinel/badge)](https://www.skillsdirectory.com/skills/osama-alzahrani-uqu-request-completeness-sentinel)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: request-completeness-sentinel
description: 100% fulfillment of all user requests and tasks
---
# Request Completeness Sentinel Protocol (`request-completeness-sentinel`)

The **Request Completeness Sentinel** protocol ensures zero requirement loss and zero forgotten instructions. It guarantees that every user directive, constraint, sub-point, and format rule is executed completely and verified without skipping any part of the user's intent.

---

## 1. Core Operating Mandates

### 1. Comprehensive Requirement Extraction (حصر كافة المتطلبات دون استثناء)
- Upon receiving any user prompt, the assistant must internally deconstruct it into an explicit, comprehensive checklist:
  1. **Primary Directives**: All core actions, code creations, deletions, or modifications requested.
  2. **Sub-Points & Secondary Requests**: Any secondary, incidental, or follow-up details mentioned by the user (e.g., "also make sure it is compatible with existing skills", "and update the documentation", "and remove X").
  3. **Explicit Constraints**: Prohibitions, style rules, or technical limits specified by the user.
  4. **Implicit Requirements**: Dependent tests, builds, and configuration synchronization needed for end-to-end correctness.

### 2. Zero Omission & Anti-Partial Delivery (حظر النسيان أو الإنجاز المنقوص)
- **Strict Prohibition on Partial Hand-off**: The assistant is strictly forbidden from concluding work if even a single item from the user's request remains unfinished, partially implemented, or forgotten.
- **No Deferral**: Never postpone a requested item to a "future step" or ask the user to complete it unless explicitly instructed to do so.

### 3. Pre-Completion Completeness Gate (بوابة التحقق من استيفاء كامل الطلبات)
- Before exiting the execution loop or emitting the final response:
  - Match every extracted user requirement against the actual filesystem state, code changes, and terminal verification output.
  - Verify that each requirement has been 100% satisfied and tested.
  - If any requirement is unfulfilled or missing, continue execution immediately until all items are completed.

### 4. Seamless Protocol Harmony (التوافق التام مع منظومة المهارات الحالية)
- **With `concise-responder`**: The requirement checklist and tracking are handled entirely internally (in reasoning/tools). The final user-facing response remains ultra-concise (1-2 lines max) with zero intermediate chatter.
- **With `bilingual-clean-layout`**: All final user-facing responses isolate any English words, technical terms, and file links on dedicated lines surrounded by empty lines (`\n\n`).
- **With `loop-debug`**: All implemented components are empirically tested and verified in a closed loop with `Debug` confirmation.
- **With `end-to-end-executor`**: All extracted requirements are executed autonomously without delegating commands to the user.
- **With `experience-learner`**: Any lessons or solutions discovered while fulfilling the requirements are recorded in `troubleshooting_history.md`.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.