AgentSkillsCN

get-directives

加载用户个人资料及 AI 个性指令。返回结构化的上下文信息,用于个性化技能行为与沟通风格。

SKILL.md
--- frontmatter
name: get-directives
description: Load user profile and AI personality directives. Returns structured context for personalizing skill behavior and communication style.
disable-model-invocation: true
allowed-tools: Read, Bash(ls)

Get Directives Sub-Skill

Loads the user profile and AI personality files from the Directives folder. Returns structured data for other skills to use for personalization.

Prerequisites

This sub-skill expects $VAULT to be set by the calling skill (via get-config).

File Locations

  • User Profile: $VAULT/00_Brain/Systemic/Directives/user-profile.md
  • AI Personality: $VAULT/00_Brain/Systemic/Directives/ai-personality.md

Execution Steps

1. Check File Existence

Check if the Directives folder and files exist:

bash
ls -la "$VAULT/00_Brain/Systemic/Directives/"

Track which files are present or missing.

2. Load User Profile (if exists)

If user-profile.md exists, read the file and extract:

From frontmatter:

  • name - Full name
  • preferred_name - How to address the user
  • role - Job title/position

From sections:

  • ## Overviewwhat_i_do
  • ## Work Contextfocus_areas, team_org, communication_style
  • ## Goals & Growthkey_goals, leadership_identity, growth_edge
  • ## Coaching Contextpatterns_to_watch, grounding_questions, success_definition

3. Load AI Personality (if exists)

If ai-personality.md exists, read the file and extract:

From frontmatter:

  • formality - casual/formal
  • directness - direct/balanced/soft
  • humor - high/medium/low/none

From sections:

  • ## Communication Style → detailed style preferences
  • ## Coaching Approachsupport_challenge, proactive_input, feedback_style
  • ## Interaction Patternsquestions_suggestions, autonomy_level, disagreement_handling

4. Return Structured Result

Return JSON output for the orchestrator to capture:

Success (both files loaded):

json
{
  "success": true,
  "user": {
    "name": "Michael Rüger",
    "preferred_name": "Michi",
    "role": "Senior Engineering Manager",
    "leadership_identity": "Coach and multiplier...",
    "growth_edge": "Delegation and trust...",
    "patterns_to_watch": ["Over-engineering", "Context switching"],
    "grounding_questions": ["What would make this week a win?"],
    "key_goals": ["Scale the team", "Improve delivery"],
    "success_definition": "Team thriving independently"
  },
  "ai": {
    "formality": "casual",
    "directness": "direct",
    "humor": "medium",
    "feedback_style": "direct",
    "proactive_input": true,
    "autonomy_level": "high"
  }
}

Partial (some files missing):

json
{
  "success": true,
  "partial": true,
  "missing": ["ai-personality.md"],
  "user": { ... },
  "ai": null,
  "message": "Directives partially loaded. Run `/init` to complete setup."
}

No files:

json
{
  "success": false,
  "error": "no_directives",
  "message": "No directives found. Using default communication style.",
  "suggestion": "Run `/init` to personalize your 2bd experience."
}

For backwards compatibility, also output human-readable summary:

code
Directives loaded successfully.

User: {preferred_name} ({role})
- Leadership Identity: {leadership_identity}
- Growth Edge: {growth_edge}
- Patterns to Watch: {patterns_to_watch}
- Grounding Questions: {grounding_questions}

AI Style: {formality}, {directness}, humor={humor}
- Feedback Style: {feedback_style}
- Autonomy: {autonomy_level}

Usage Guide for Parent Skills

How to Apply User Profile

FieldUse For
preferred_nameAddress the user by name ("Good morning, Michi!")
leadership_identityFrame leadership intentions and suggestions
growth_edgeGenerate growth-focused coaching prompts
patterns_to_watchProactive reminders when patterns might emerge
grounding_questionsInclude in reflection/insight sections
key_goalsConnect daily work to larger objectives
success_definitionFrame wins and progress

How to Apply AI Personality

FieldEffect
formality: casualConversational tone, no corporate speak
formality: formalProfessional, structured communication
directness: directSay what needs to be said, no softening
directness: balancedDirect but considerate
humor: highInclude humor, make it enjoyable
humor: low/noneKeep it professional, minimal humor
feedback_style: directGive feedback straight, no sandwich method
proactive_input: yesSuggest and advise without being asked
autonomy_levelHow much to check in vs. execute independently

Graceful Degradation

If directives aren't loaded:

  • Use neutral, professional tone as default
  • Skip personalized coaching references
  • At the end of the skill, suggest running /init

Difference from Coaching Context

SourcePurposeStability
DirectivesWHO the user is, HOW to communicateStable (rarely changes)
Year.md/Quarter.mdWHAT they're focused on nowTemporal (changes by period)

Skills should load BOTH:

  1. Directives first (identity and style)
  2. Then temporal coaching context (current goals and focus)