Skip to content
Back to skills

Onboard Report

ASecurity

Department head / company viewer view of employee onboarding on the total-agent-memory team server. Use when the user types /onboard-report, asks how new employees are doing with training, who finished which module, quiz scores, wants to grade open quiz answers, or to build/edit the department curriculum or quizzes.

  • 72 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 29, 2026
ai-agents

Works with

  • cli

Security analysis

A100/100

Scanned September 29, 2026

npx -y skills add vbcherepanov/total-agent-memory --skill onboard-report --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Onboard Report?

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

Security grade badge for Onboard Report
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vbcherepanov-onboard-report/badge)](https://www.skillsdirectory.com/skills/vbcherepanov-onboard-report)

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

SKILL.md
---
name: onboard-report
description: >
  Department head / company viewer view of employee onboarding on the
  total-agent-memory team server. Use when the user types /onboard-report, asks how
  new employees are doing with training, who finished which module, quiz scores,
  wants to grade open quiz answers, or to build/edit the department curriculum or quizzes.
argument-hint: "[department id]"
---

# /onboard-report — onboarding results and curriculum

Tools are named `onboarding_*` (possibly client-prefixed). The server enforces
access: department heads (team role `manager`) see and manage their own
department; `company_viewer` and `superadmin` see every department; `superadmin`
may also manage. A `Forbidden` error means the user lacks that role — say so.

## 1. Pick the department

Use `$ARGUMENTS` if given. Otherwise call `onboarding_overview` and choose from
`viewable_teams` (report) or `manageable_teams` (editing).

## 2. Report

Call `onboarding_team_report(team_id)` and present:

- a table: employee × module with status (not started / in progress /
  quiz available / awaiting review / passed / failed), lessons done, best quiz
  score, and the relevant date (started, finished, passed);
- who is stuck (failed with no attempts left, or no activity) and lessons
  flagged as updated since people studied them (`summary.updated_lessons`);
- the latest entries of `log` in plain language.

For one person's detail: `onboarding_progress(team_id, user_id)`.

## 3. Grade open answers (department head)

For each item in `grading_queue`, show prompt, rubric and answer, ask the head
for points (0..`points`) and an optional comment, then call
`onboarding_grade(attempt_id, question_id, score, comment)`. Grading the last
open answer finalizes the attempt. Never grade without the head's decision.

## 4. Build or edit the curriculum (department head)

1. `onboarding_curriculum_get(team_id)` → current curriculum and `revision`.
2. For a first version, `onboarding_curriculum_draft(team_id)` groups the
   department's memory into modules/lessons (LLM-polished only if the server
   has one). Review it with the head: reorder, rename, remove, add lesson text.
3. Save with `onboarding_curriculum_set(team_id, expected_revision, curriculum)`.
   Keep existing `id`s so employees keep their progress; omitted items are archived.
4. Quizzes per module: `onboarding_quiz_draft(module_id)` (needs a server LLM;
   drafts are checked against lesson quotes) or write questions with the head,
   then `onboarding_quiz_set(module_id, expected_revision, questions)`.
   Question types: `single`, `multiple` (zero-based `correct` indexes), `open`
   (needs `rubric`). Set `pass_threshold` and `max_attempts` per module in the
   curriculum.

Always show the head what will be saved before calling a `*_set` tool.

Files in this skill

  • SKILL.md2.8 KB
  • agents/openai.yaml324 B

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.