Bootstrap skill — skill routing, gate enforcement, mandatory workflows
From mbscodenpx claudepluginhub mbstools/mbscode --plugin mbscodeThis skill uses the workspace's default tool permissions.
INIT.mdSearches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.
Guides implementation of event-driven hooks in Claude Code plugins using prompt-based validation and bash commands for PreToolUse, Stop, and session events.
| Skill | Trigger |
|---|---|
brainstorming | Before creative work (building, designing, features) |
writing-plans | Spec ready, before coding |
executing-plans | Executing a plan inline |
subagent-driven-development | Plan with independent parallelizable tasks |
autonomous-review | After milestone or task group |
verification-before-completion | Before claiming work is done |
systematic-debugging | Bugs, failures, unexpected behavior |
finishing-a-development-branch | All tasks complete, verification passes |
project-documentation | Project structure changes |
test-driven-development | Tests for bug fixes, parsers, business logic |
writing-skills | Creating or modifying skills |
This skill (using-mbscode) is the bootstrap — it is not listed above because it loads automatically.
Access: Claude Code → Skill tool by name (e.g., mbscode:brainstorming). Other platforms → read skills/<name>/SKILL.md from plugin root.
Custom skills: Define in PROJECT.md ## Custom Skills. Same-name overrides built-in. Author guide: skills/writing-skills/SKILL.md.
Invoke matching skills BEFORE any response. If unsure, invoke — skip if wrong.
HARD-GATE: No GATES.md at project root → read skills/using-mbscode/INIT.md → initialize first.
Pre-task: verify all Pre-Task Gates. Post-task: verify all Post-Task Gates with evidence. Gate fails → fix → re-check. Two failures → change approach. Three → escalate to user.
| Missing | Recovery |
|---|---|
SESSION.md | Create fresh, no active work |
PROJECT.md | Re-run init for this file, ask user |
| Skill file | Warn, skip |
LESSONS_LEARNED.md | Create with # Lessons Learned header |
| Type | Skill | Plan? |
|---|---|---|
| Trivial (< 3 steps, single file) | None (gate check only) | No |
| Bug ("X is broken") | systematic-debugging | No |
| Feature (3+ steps) | brainstorming → plan | Yes |
| Refactor (3+ steps) | brainstorming or writing-plans | Yes |
| Docs only | project-documentation | No |
| Mixed | Split or sequential (bug first) | Depends |
| Hotfix (prod down) | systematic-debugging, branch hotfix/<issue> | No |
Unclear if trivial → treat as feature → brainstorming.
Priority: existing code patterns → PROJECT.md conventions → language idioms → community standard.
Same issue twice after fix → change approach. Three → escalate. Applies everywhere.
Never echo, log, or commit secrets. Store in env vars.
Rigid (verification-before-completion, systematic-debugging, executing-plans, test-driven-development): Follow exactly. Flexible (brainstorming, project-documentation, writing-plans, writing-skills): Adapt to context. Mixed (subagent-driven-development, autonomous-review, finishing-a-development-branch): Follow structure, adapt detail.