{/* Generated by scripts/public_surface/generate_capability_catalog.py. Do not edit by hand. */}
Genesis maps 50 user-facing capabilities to a command, in 9 families. 13 further modules are internal plumbing with no command of their own.
This count is different from the number of registered engines on the home page. Engines are the units the decision engine schedules; capabilities are the things a person can ask for. Several capabilities run inside one engine, and some engines serve several capabilities.
| Column |
Meaning |
| Maturity |
stable: default install, covered by tests. optional: needs an extra or an external tool. experimental: no direct test coverage yet |
| Writes |
What it can leave behind. .genesis/ and reports stays inside the project’s own state folder and generated reports; files on disk is always opt-in or a dry run first |
| Needs |
Network, LLM and optional dependencies |
A row with Writes: nothing does not change your files. Commands that write state, such as recover, harden and decide, put it under .genesis/ and in generated reports. The machine-readable source is data/public-capabilities.json.
| Capability |
What it does |
Command |
Maturity |
Writes |
Needs |
| Research Orchestrator |
Merges four research streams, ranks them, enforces the research floor Live search needs network; –json-data runs offline on supplied results. |
genesis research "<topic>" [--json-data FILE] |
stable |
.genesis/ and reports |
offline by default |
| Pitfall Ranker |
Dedupes and scores pitfalls; ranks without needing engagement counts |
genesis research "<topic>" --json-data FILE |
stable |
nothing |
offline |
| Media Research |
Builds YouTube/Reddit/Instagram queries and parses the results |
genesis research "<topic>" --video |
optional |
.genesis/ and reports |
needs network; yt-dlp, ffmpeg |
| Video to Pitfall |
Turns a /watch analysis into cited PITFALLS.md entries |
genesis research "<topic>" --absorb FILE |
optional |
.genesis/ and reports |
offline; ffmpeg |
| Research Outline |
States what to research before investigating: an items x fields grid, saved as outline.json/fields.json |
genesis decide "research <topic>" |
stable |
.genesis/ and reports |
offline |
| Field Intelligence |
Practitioner signal from Reddit Answers and similar surfaces |
genesis decide "research <topic>" |
stable |
.genesis/ and reports |
offline |
| Source Registry |
Which sources are trusted for which kind of claim |
genesis decide "research <topic>" |
stable |
.genesis/ and reports |
offline |
| Source Anchor |
Pins every claim to a retrievable source |
genesis decide "research <topic>" |
stable |
nothing |
offline |
| Evidence Pack |
Bundles the evidence behind a decision for review |
genesis decide "research <topic>" |
stable |
.genesis/ and reports |
offline |
| Product Intelligence |
Competitive and positioning signal for the vision Telemetry is opt-in and off by default. |
genesis decide "research <topic>" |
stable |
.genesis/ and reports |
offline |
| Capability |
What it does |
Command |
Maturity |
Writes |
Needs |
| Recovery Scan |
Diagnoses an existing codebase: drift, broken imports, decay |
genesis recover [PATH] |
stable |
.genesis/ and reports |
offline |
| Recovery Report |
Renders the recovery scan as a prioritised report |
genesis recover [PATH] |
stable |
.genesis/ and reports |
offline |
| Fragility Classifier |
STABLE / FRAGILE / VOLATILE per module, from git churn + coverage |
genesis recover [PATH] |
stable |
.genesis/ and reports |
offline |
| Drift Detector |
Detects divergence between the plan and the code on disk |
genesis recover [PATH] |
stable |
nothing |
offline |
| Drift Scorer |
Scores how far the project has drifted, with a trend |
genesis recover [PATH] |
stable |
nothing |
offline |
| Decay Regressor |
Projects where quality is heading if nothing changes |
genesis recover [PATH] |
stable |
nothing |
offline |
| Git Analyzer |
Churn, hotspots, and authorship signals from git history |
genesis recover [PATH] |
stable |
nothing |
offline; git |
| Capability |
What it does |
Command |
Maturity |
Writes |
Needs |
| Import Graph |
Multi-language dependency graph with cycle detection |
genesis recover [PATH] |
stable |
.genesis/ and reports |
offline |
| Import Audit |
Finds imports that resolve to nothing |
genesis recover [PATH] |
stable |
nothing |
offline |
| Architecture Scorer |
0-100 quality score across 4 dimensions, 6 adaptive profiles |
genesis recover [PATH] |
stable |
.genesis/ and reports |
offline |
| Anti-Pattern Detector |
7 structural detectors: god-class, hub-file, circular deps, dead code |
genesis recover [PATH] |
stable |
nothing |
offline |
| Refactoring Planner |
Tier-1/2 refactor steps with projected score impact |
genesis decide "refactor <target>" |
stable |
nothing |
offline |
| C4 Generator |
C4 Level 1-3 architecture diagrams as GitHub-native Mermaid |
genesis decide "document the architecture" |
stable |
nothing |
offline |
| Knowledge Graph |
Links code, CVEs, risks, and decisions into one queryable graph |
genesis decide "diagnose the project" |
stable |
.genesis/ and reports |
offline |
| Capability |
What it does |
Command |
Maturity |
Writes |
Needs |
| Decision Engine |
Routes an instruction to the right engines and gates the writes |
genesis decide "<instruction>" |
stable |
.genesis/ and reports |
offline |
| Red Team Critic |
Attacks the plan and separates evidence from inference Deterministic checks always run; the LLM pass is additive and disclosed when skipped. |
genesis decide "<instruction>" |
stable |
.genesis/ and reports |
offline by default; LLM optional; anthropic (committee extra) |
| Engine Bootstrap |
Registers every production engine, in dependency order |
genesis decide "<instruction>" |
stable |
nothing |
offline |
| Rules Engine |
Hard and soft project rules, enforced at the gate |
genesis decide "<instruction>" |
stable |
nothing |
offline |
| Intent Classifier |
Maps a plain-English instruction onto a mode |
genesis decide --classify-only "<instruction>" |
stable |
nothing |
offline |
| Capability |
What it does |
Command |
Maturity |
Writes |
Needs |
| Security Templates |
STRIDE threat model + OWASP Top 10 checklist, archetype-aware Generates a checklist and threat model; it is not a vulnerability scanner. |
genesis harden [PATH] |
stable |
.genesis/ and reports |
offline |
| Secrets Scanner |
Finds committed credentials and tokens |
genesis harden [PATH] |
stable |
nothing |
offline |
| Dependency Scanner |
Maps third-party imports per module and looks up their CVEs |
genesis deps [PATH] |
stable |
nothing |
offline |
| Package Registry |
Validates packages against PyPI/npm/crates.io/Maven/NuGet + OSV |
genesis deps [PATH] --package NAME |
stable |
nothing |
offline by default |
| Dependency Index |
Caches the resolved dependency graph between runs |
genesis deps [PATH] |
stable |
nothing |
offline |
| Stdlib Filter |
Separates standard-library imports from third-party ones |
genesis deps [PATH] |
stable |
nothing |
offline |
| Capability |
What it does |
Command |
Maturity |
Writes |
Needs |
| Cross-Session Memory |
Restores research/build context from a previous session |
genesis memory --sessions |
stable |
.genesis/ and reports |
offline |
| Project Memory |
The per-project .genesis/*.md memory files |
genesis memory |
stable |
.genesis/ and reports |
offline |
| Learning Engine |
Learns from accepted and rejected findings over time |
genesis telemetry |
stable |
.genesis/ and reports |
offline |
| Progress Report |
What changed since the last session |
genesis memory --status |
stable |
.genesis/ and reports |
offline |
| Capability |
What it does |
Command |
Maturity |
Writes |
Needs |
| Autonomous Sync |
Gate + findings + auto-apply for the GREEN zone Auto-applies only the GREEN zone; –dry-run reports without writing. |
genesis sync |
stable |
.genesis/ and reports |
offline |
| Skill Fetcher |
Fetches a whitelisted skill pack into an ephemeral sandbox Clones into a short-lived sandbox (default and maximum 2 hours). |
genesis fetch [SOURCE] |
stable |
files on disk (opt-in or dry run first) |
needs network; git |
| Auto-Purge |
Reports and removes expired ephemeral resources Dry run by default; –apply removes expired worktrees and locks. |
genesis purge [--apply] |
stable |
files on disk (opt-in or dry run first) |
offline |
| Readiness Check |
Install and optional-dependency readiness |
genesis doctor |
stable |
nothing |
offline |
| Capability |
What it does |
Command |
Maturity |
Writes |
Needs |
| Companion UI |
The Genesis health page served locally |
genesis companion |
stable |
.genesis/ and reports |
offline |
| Companion Window |
The Floating Assistant shell Native floating window; falls back to the browser page when pywebview is unavailable. |
genesis companion --ui |
experimental |
.genesis/ and reports |
offline; pywebview |
| Companion Backend |
WebSocket + IDE bridge backing the Companion Local health page uses the standard library only. |
genesis companion --serve |
optional |
.genesis/ and reports |
offline; plyer (notifications) |
| Canvas Workspace |
Self-contained HTML workspace for the current project |
genesis ui |
stable |
.genesis/ and reports |
offline |
| Model Store |
Local STT/TTS model download and cache –setup downloads local STT/TTS models into ~/.genesis/models. |
genesis companion --setup |
optional |
.genesis/ and reports |
needs network; voice extra |
| Capability |
What it does |
Command |
Maturity |
Writes |
Needs |
| MCP Advisor |
Recommends MCP servers and skills for this project Recommends MCP servers and skills; installs nothing. |
genesis advise |
stable |
nothing |
offline |
| MCP Tools |
Exposes Genesis engines as MCP tools Exposes four read-only tools over MCP. |
genesis advise |
stable |
nothing |
offline |
| Command |
Status |
What it does |
genesis init |
PUBLIC |
Research GitHub, mine pitfalls, scaffold a new project (needs an LLM key or a local Ollama model) |
genesis research |
PUBLIC |
Multi-source research on a topic |
genesis resolve |
ADVANCED |
Look up a known solution: local vault first, then Stack Overflow |
genesis publish |
PUBLIC |
Draft a Show HN post and release notes with an LLM (needs LLM_API_KEY); you post them yourself |
genesis config |
PUBLIC |
Set, get or show API keys |
genesis upgrade |
LEGACY |
Kept for compatibility; Genesis has no paid tier |
genesis decide |
PUBLIC |
Run a full decision session on an instruction |
genesis explain |
PUBLIC |
Print the last session’s decision log |
genesis memory |
PUBLIC |
Show or manage per-project memory in .genesis/ |
genesis engines |
PUBLIC |
List every engine and the command that reaches it |
genesis deps |
PUBLIC |
Third-party dependencies per module, plus advisories |
genesis ui |
PUBLIC |
Generate the self-contained HTML canvas workspace |
genesis companion |
ADVANCED |
Local health page, notifications and optional voice |
genesis sync |
ADVANCED |
Autonomous sync: gate, findings, GREEN-zone auto-apply |
genesis doctor |
PUBLIC |
Readiness check for the install |
genesis recover |
PUBLIC |
Diagnose drift, broken imports, anti-patterns and decay |
genesis harden |
PUBLIC |
STRIDE threat model, OWASP Top 10 checklist, secrets scan |
genesis advise |
PUBLIC |
Recommend MCP servers and skills for this project |
genesis fetch |
ADVANCED |
Fetch a trusted skill pack into a short-lived sandbox |
genesis purge |
ADVANCED |
Remove expired worktrees and locks (dry run by default) |
genesis telemetry |
ADVANCED |
Manage opt-in telemetry; the default is off |
genesis license |
LEGACY |
Kept for compatibility; no license key is needed |
PUBLIC is the everyday surface. ADVANCED is supported but narrower. LEGACY is kept so old scripts keep working and adds nothing new.