epci971

component-advisor

Passive detection of EPCI component creation opportunities. Identifies repetitive patterns that could become skills, commands or subagents. Use when: workflow analysis, recurring pattern detection. Not for: active component creation (use /epci:create).

epci971 0 Updated 8mo ago

Resources

1
GitHub

Install

npx skillscat add epci971/claude-epci/component-advisor

Install via the SkillsCat registry.

SKILL.md

Component Advisor

Overview

Passive skill that detects opportunities for new EPCI component creation
based on usage pattern analysis.

Reference Documents:

Automatic Detection

New Skill Indicators

Signal Score Example
Pattern repeated 3+ times +3 Same validation in multiple commands
Technical domain not covered +2 New unsupported stack
Documentation frequently consulted +2 Repeated searches on same topic
Copy-paste of guidelines +1 Same conventions applied

Suggestion threshold: Score ≥ 4

New Command Indicators

Signal Score Example
Repeated action sequence +3 Same recurring manual workflow
Frequent skill combination +2 Always the same skills together
Documented but not automated process +2 Guide followed manually
Explicit user request +3 "I'd like a command for..."

Suggestion threshold: Score ≥ 4

New Subagent Indicators

Signal Score Example
Repeated specialized validation +3 Specific security check
Recurring manual review +2 Same checklist applied
Narrow domain expertise +2 Specialized knowledge required
Standardized report format +1 Same report structure

Suggestion threshold: Score ≥ 4

Suggestion Format

When a threshold is reached:

💡 **COMPONENT OPPORTUNITY DETECTED**

### Suggested Type: [Skill | Command | Subagent]

**Identified pattern:**
[Description of detected pattern]

**Occurrences:**
- [Occurrence 1]
- [Occurrence 2]
- [Occurrence 3]

**Estimated benefits:**
- [Benefit 1]
- [Benefit 2]

**Proposal:**

/epci:create [type] [suggested-name]


**Confidence score:** [X/10]

---
*Automatic suggestion - Ignore if not relevant*

Monitored Patterns

For Skills

Pattern Potential Domain
Repeated searches on a tech New stack skill
Manually applied conventions Conventions skill
Often cited best practices Patterns skill
Frequently used external tools Integration skill

For Commands

Pattern Potential Command
Repeated tool sequence Composite command
Manual multi-step workflow Automation command
Process with breakpoints Structured command
Action + validation + report Workflow command

For Subagents

Pattern Potential Subagent
Recurring validation Validator agent
Specialized analysis Analyzer agent
Review with checklist Reviewer agent
Formatted generation Generator agent

Configuration

Enable/Disable

The component-advisor is passive by default.
It observes and suggests without interrupting workflow.

Customizable Thresholds

component_advisor:
  skill_threshold: 4
  command_threshold: 4
  subagent_threshold: 4
  suggestion_frequency: "on_pattern_detected"  # or "end_of_session"

Detection Examples

Example 1: New Skill Detected

💡 COMPONENT OPPORTUNITY: Skill

Pattern: Kubernetes documentation consulted 5 times
         Same deployment structure applied 3 times

Suggestion: /epci:create skill kubernetes-patterns

Benefits:
- Auto-detection of K8s projects
- Standardized patterns
- Reduced search time

Example 2: New Command Detected

💡 COMPONENT OPPORTUNITY: Command

Pattern: Repeated sequence
         1. Lint → 2. Test → 3. Build → 4. Deploy

Suggestion: /epci:create command ci-pipeline

Benefits:
- Process automation
- Consistency across projects
- Time savings

Example 3: New Subagent Detected

💡 COMPONENT OPPORTUNITY: Subagent

Pattern: Accessibility checklist applied 4 times
         Same report format generated

Suggestion: /epci:create agent a11y-auditor

Benefits:
- Automatic audit
- Standardized report
- No missed criteria

Metrics

Metric Description
Patterns detected Number of patterns identified
Suggestions issued Number of suggestions proposed
Suggestions accepted Components actually created
Adoption rate % suggestions → components

Limitations

  • Detection based on current session
  • No memory between sessions (unless context provided)
  • Suggestions are indicative, not prescriptive
  • Requires repeated patterns to detect