auto-skill-bootstrap

$npx mdskill add SteelMorgan/1c-agent-based-dev-framework/auto-skill-bootstrap

Automatically inventories project skills, detects missing capabilities, and installs trusted skills to fill gaps.

  • Helps developers manage skill coverage in projects by identifying and addressing missing functionalities.
  • Integrates with Skills CLI to search for and install skills from skills.sh under a trust policy.
  • Uses skills-manifest.json and state.json to ensure deterministic, idempotent recommendations across changing requirements.
  • Presents a structured install plan and optionally installs skills after user confirmation for safety.

SKILL.md

.github/skills/auto-skill-bootstrapView on GitHub ↗
---
name: auto-skill-bootstrap
description: Deterministic helper to inventory existing project skills, detect missing capability coverage, search skills.sh via Skills CLI, and (optionally) install missing skills under a trust policy. Uses skills-manifest.json + state.json to stay idempotent across changing requirements.
---

# Auto Skill Bootstrap (project-level)

## What this is

This skill defines a **repeatable** workflow for:

- indexing already-present project skills into `skills-manifest.json`
- mapping the user request to **capabilities**
- finding gaps (capabilities not covered by existing skills)
- searching Skills CLI (`npx skills find ...`) for candidate skills
- filtering candidates by **trust policy**
- producing a short, structured “install plan”
- optionally installing skills (only under strict conditions)

The heavy lifting is done by deterministic scripts shipped with this skill.

## Files (state + inventory)

- Inventory of skills (generated): `.cursor/skills/skills-manifest.json`
- Bootstrap state (generated): `.cursor/skills/auto-skill-bootstrap/state.json`

## Hard rules

- **Never** install skills from unknown sources automatically.
- If multiple plausible candidates exist for a capability, **ask the user** to choose (multi-select).
- After any install/remove/update of skills, regenerate `skills-manifest.json` (do not edit JSON by hand).

## Deterministic commands

### 1) Rebuild skills manifest

Run inside the repo root:

```bash
python3 .cursor/skills/auto-skill-bootstrap/bin/update-manifest.py
```

### 2) Search for missing capability skills (no install)

```bash
python3 .cursor/skills/auto-skill-bootstrap/bin/auto-skill-bootstrap.py \
  --cap docker github devcontainers \
  --no-install
```

Outputs:
- `.cursor/skills/auto-skill-bootstrap/candidates.json` (grouped by capability)
- updates `.cursor/skills/auto-skill-bootstrap/state.json`

### 3) Optional: install (trust-policy only)

Only when **explicitly allowed** by the user or project policy.

```bash
python3 .cursor/skills/auto-skill-bootstrap/bin/auto-skill-bootstrap.py \
  --cap docker github \
  --install-allowlisted \
  --max-per-cap 1
```

## Trust policy

Trust rules live here and are deterministic:

- `.cursor/skills/auto-skill-bootstrap/trust-policy.json`

Default stance: allowlist only.

## When to run

- At the start of a new task (before deep work).
- Again when the user introduces new constraints/tech (“also add CI”, “needs k8s”, etc.).

More from SteelMorgan/1c-agent-based-dev-framework

SkillDescription
1c-ai-agent-cliCLI 1C BSL Agent Framework — tools/install.py (clone, install). Используй при клонировании репозитория, установке компонентов в проект, настройке IDE (Cursor, Claude Code, Windsurf, VS Code+Continue).
agent-debugПаттерн отладочных сообщений для 1С BSL. Используй, когда стандартная диагностика (event-log, скриншоты) не даёт понять фактическое поведение системы — нужно вставить временные точки логирования в код, запустить тест и проанализировать записи ЖР.
agent-developmentCreate custom subagents for specialized AI tasks. Use when the user wants to create a new type of subagent, set up task-specific agents, configure code reviewers, debuggers, or domain-specific assistants with custom prompts.
agent-development-ext>
agent-git-workflowStandardizes git workflow for the AI agent in the sandbox devcontainer: work in agent/<task>-<yyyymmdd>, integrate via agent, never push to main/master, open PRs via GitHub CLI. Use when the user asks to create branches, push changes, open PRs, or follow this sandbox repo setup.
code-navigationНавигация по коду (Code Navigation). Навык учит агента **эффективно перемещаться по BSL-коду** с помощью LSP (Language Server Protocol).
config-operationsОперации с конфигурацией 1С (CF) — init, info, edit, validate. Используй при создании конфигурации, анализе структуры, изменении свойств и ChildObjects, валидации Configuration.xml.
epf-buildСобрать EPF/ERF из XML-исходников. Используй после внесения временных диагностических правок в разобранную обработку.
epf-dumpРазобрать EPF/ERF в XML-исходники. Используй, когда нужно быстро получить исходный код внешней обработки или отчета для анализа и временной модификации.
epf-operationsОперации с внешними обработками 1С (EPF) — создание, добавление форм и шаблонов. Используй при epf init, add-form, add-template, add-attribute, add-tabular-section.