Documentation > Features > Feature Reference
Complete reference of all skill-creator features organized by subsystem. Each feature includes its module location, key capabilities, and version introduced.
Core Skill Management
| Feature | Module | Version | Description |
|---|
| Skill CRUD | src/storage/skill-store.ts | v1.0 | Create, read, update, delete skills with metadata validation and scope resolution (project vs user level) |
| Skill Index | src/storage/skill-index.ts | v1.0 | In-memory full-text index for fast matching by name, description, triggers, and tools |
| Skill Validation | src/validation/skill-validation.ts | v1.1 | 15 validators covering safety, budget, compatibility, path traversal, reserved names, YAML safety |
| Conflict Detection | src/conflicts/ | v1.2 | Semantic similarity analysis to detect overlapping or contradictory skills before they cause issues |
| Skill Format | src/types/skill.ts | v1.0 | Claude Code official SKILL.md format with gsd-skill-creator extensions for triggers and learning metadata |
Observation and Detection
| Feature | Module | Version | Description |
|---|
| Session Observer | src/observation/session-observer.ts | v1.4 | Captures tool sequences, file patterns, command frequencies as append-only JSONL observations |
| Pattern Analyzer | src/detection/pattern-analyzer.ts | v1.4 | DBSCAN clustering to identify recurring patterns across sessions with configurable thresholds |
| Skill Generator | src/detection/skill-generator.ts | v1.5 | Transforms detected patterns into complete SKILL.md scaffolds with safety validation and progressive disclosure |
| Pattern Discovery CLI | src/cli/discover.ts | v1.6 | Manual pattern scan: npx skill-creator discover --threshold 2 --since 7d --incremental |
Learning and Refinement
| Feature | Module | Version | Description |
|---|
| Feedback Capture | src/learning/feedback-store.ts | v1.8 | Three feedback types (correction, override, rating) with diff analysis and session correlation |
| Refinement Engine | src/learning/refinement-engine.ts | v1.9 | Bounded refinement: 3+ corrections required, 7-day cooldown, max 20% content change, always requires user approval |
| Drift Tracker | src/learning/drift-tracker.ts | v1.10 | Monitors skill output consistency over time, detects concept drift, format drift, and regression |
| Contradiction Detector | src/learning/contradiction-detector.ts | v1.10 | Identifies conflicting feedback patterns that signal skill ambiguity |
Activation Pipeline
| Feature | Module | Version | Description |
|---|
| 6-Stage Pipeline | src/application/skill-pipeline.ts | v1.8 | Score, Resolve, ModelFilter, CacheOrder, Budget, Load — balances relevance against token budget |
| Token Budgeting | src/validation/budget-validation.ts | v1.8 | Enforces 2-5% context window allocation with deferred loading for overflow skills |
| Model Profiles | src/application/model-profiles.ts | v1.11 | Three profiles (quality/balanced/budget) controlling model assignment per agent role |
| Skill Inheritance | src/composition/skill-resolver.ts | v1.11 | Skills can extend other skills; resolver merges parent before child with separator |
Agent and Team Composition
| Feature | Module | Version | Description |
|---|
| Co-Activation Tracker | src/agents/co-activation.ts | v1.5 | Records which skills activate together; suggests agent composition after 5+ co-activations over 7+ days |
| Agent Generator | src/agents/agent-generator.ts | v1.6 | Creates multi-agent definitions from detected skill clusters with model and tool assignment |
| Team Templates | src/teams/templates/ | v1.7 | 5 topologies: leader-worker, pipeline, swarm, router, map-reduce with GSD specializations |
| Team Validator | src/teams/team-validator.ts | v1.7 | Validates member agents, detects task cycles, tool overlap, skill conflicts, role coherence |
Safety and Security
| Feature | Module | Version | Description |
|---|
| Generation Safety | src/validation/generation-safety.ts | v1.9 | Dangerous command detection (rm -rf, git reset --hard, etc.) during skill generation |
| Path Safety | src/validation/path-safety.ts | v1.9 | Prevents path traversal in skill names and file operations |
| YAML Safety | src/validation/yaml-safety.ts | v1.9 | Safe YAML parsing only — no arbitrary code execution from deserialization |
| JSONL Safety | src/validation/jsonl-safety.ts | v1.10 | Validates append-only observation files against tampering and injection |
| Audit Logger | src/safety/audit-logger.ts | v1.10 | Logs all skill mutations for compliance and forensic review |
| File Lock | src/safety/file-lock.ts | v1.10 | Prevents concurrent skill modifications from parallel agents |
| Operation Cooldown | src/safety/operation-cooldown.ts | v1.10 | Rate limits risky operations (skill creation, deletion, bulk changes) |
GSD Integration
| Feature | Module | Version | Description |
|---|
| GSD Orchestrator | src/orchestrator/ | v1.7 | 5-stage intent classification: exact match, lifecycle filter, Bayesian, semantic fallback, confidence resolution |
| Phase Skill Loading | src/orchestrator/skill-loader.ts | v1.11 | Loads relevant skills before GSD phase execution; includes skills in forked subagent contexts |
| GSD Reference Injection | src/detection/gsd-reference.ts | v1.12 | Embeds phase goals, requirement IDs, and related file paths into generated skills |
| Dashboard Generation | desktop/src/dashboard/ | v1.13 | 6 HTML dashboard pages (index, requirements, roadmap, milestones, state, console) from .planning/ data |
Desktop (GSD-OS)
| Feature | Module | Version | Description |
|---|
| WebGL CRT Engine | desktop/src/engine/ | v1.18 | GPU-accelerated rendering with scan lines, bloom, curvature, phosphor glow — Amiga Workbench aesthetic |
| Calibration Wizard | desktop/src/calibration/ | v1.19 | First-boot experience: CRT settings, color palette selection, theme mode, accessibility detection |
| Window Manager | desktop/src/wm/ | v1.20 | Amiga-style window management with titlebar gadgets, drag regions, depth ordering, icon area |
| Desktop Shell | desktop/src/shell/ | v1.21 | Icons, taskbar, system menu, keyboard navigation, process monitor — complete desktop environment |
| Tauri Integration | desktop/src-tauri/ | v1.21 | Tauri v2 native wrapper with IPC commands, file watcher, PTY terminal, Rust backend |
Specialized Systems (v1.22+)
| Feature | Module | Version | Description |
|---|
| Electronics Education | src/electronics/ | v1.29 | MNA circuit simulator (DC/AC/transient), digital logic simulator (8 gate types), Safety Warden (3 modes: annotate/gate/redirect), 15 modules, 77 labs, 5 specialized engines (DSP, GPIO, PLC, Solar, PCB). 10,707 tests. |
| Brainstorm Sessions | src/brainstorm/ | v1.32 | 16 techniques across 4 categories (individual, collaborative, analytical, visual). 8 agents (Facilitator, Ideator, Questioner, Analyst, Mapper, Persona, Critic, Scribe). 5 pathways. 4-loop session bus. Critic architecturally gated to Converge phase — Osborn's rules enforced at code level. |
| Vision to Mission | src/vtm/ | v1.30 | Automated vision-to-mission pipeline: document parsing, archetype classification, research compilation, wave planning with parallel track detection, model assignment engine (60/40 principle), cache optimization. |
| MCP Integration | src/mcp/ | v1.31 | Full MCP stack: gateway server (Streamable HTTP, bearer auth, per-tool scopes), host manager (stdio transport, crash recovery, exponential backoff), agent bridge (any agent as MCP server), security pipeline (SHA-256 hash gates, trust lifecycle with 30-day decay), template generators, trace panel. |
| OpenStack Cloud | src/openstack/ | v1.33 | 19 skills for 8 OpenStack services. 31 agents across 3 crews (Deployment, Operations, Documentation). 9 communication loops. 118 pre-deploy validation checks. NASA SE methodology: 7-chapter sysadmin guide, 80 procedures, 44 runbooks, 55-requirement V&V plan. |
| Chipset Architecture | src/chipset/ | v1.28 | 10-position Den staff with token budgets (total: 0.59). Positions: coordinator, planner, executor, verifier, relay, monitor, dispatcher, configurator, analyst, observer. Configuration-driven multi-agent orchestration. |
CLI Commands
# Core operations
npx skill-creator start # Initialize project
npx skill-creator status --json # Budget, loaded skills, projections
npx skill-creator create # Create skill from observation
npx skill-creator validate # Validate skill metadata
npx skill-creator test run # Run activation simulation
npx skill-creator calibrate # Tune activation thresholds
# Discovery and analysis
npx skill-creator discover # Scan sessions for patterns
npx skill-creator detect-conflicts # Find overlapping skills
npx skill-creator benchmark # Run precision/recall benchmark
# Teams and agents
npx skill-creator team create # Create team configuration
npx skill-creator team validate # Validate team members and topology
npx skill-creator agent suggest # Review co-activation clusters
# GSD orchestration
npx skill-creator orchestrator discover # List GSD commands
npx skill-creator orchestrator classify # Classify user intent
npx skill-creator orchestrator lifecycle # Suggest next action
By the Numbers (v1.33)
| Metric | Count |
|---|
| Total versions | 33 |
| Total phases | 325 |
| Total tests | 12,545 |
| Approximate LOC | ~410,000 |
| Source modules | 16+ subsystems |
| Skills installed | 15 project-level |
| GSD commands | 20+ |
| Brainstorm techniques | 16 |
| Electronics labs | 77 |
| OpenStack agents | 31 |
| OpenStack runbooks | 44 |
