Elin, S. M. (2026). Terminal Wraith: A local-first agentic-engineering environment for code, reasoning, and verification.

Project Card: Terminal Wraith is a local-first agentic-engineering environment for bounded, inspectable, and verified AI-assisted work. It organizes model output through the Recursive Daemon, a control loop that observes, plans, acts, verifies, confesses, revises, and converges only under explicit constraints. The system combines verification profiles, approval gates, secret scanning, threat modeling, cryptographic review, resource budgets, governed memory, model-architecture awareness, provenance logs, and human acceptance. Its guiding principle is simple: plausible output is not verified work. Its motto, Ex umbra ad evidentiam—from shadow to evidence—captures its purpose: to bring hidden assumptions, uncertain outputs, unsafe actions, stale memory, and opaque failures into view through trace, test, review, and verification.

Terminal Wraith Repository Map

Terminal Wraith is packaged as a local-first AI coding and reasoning environment. The repository is organized around the launcher, the Recursive Daemon, verification tools, security checks, reasoning modules, and project documentation.

Structure

terminal-wraith/
├── README.md
├── PROJECT_PAGE.md
├── QUICKSTART.md
├── INSTALL.md
├── REPOSITORY_MAP.md
├── CODE_STYLE_AUDIT.md
├── LICENSE_NOTICE.md
├── .gitignore
├── .gitattributes
├── Install-TerminalWraith.ps1
├── tools/
│   └── check_code_style.py
├── docs/
│   ├── Terminal_Wraith_User_Setup_Guide.md
│   └── Terminal_Wraith_User_Setup_Guide.pdf
├── website/
│   ├── project-page.html
│   └── repository-map.html
└── TerminalWraith/
    ├── terminal-wraith.ps1
    ├── model_config.json
    ├── shared_model_client.py
    ├── shared_model_config.py
    ├── Update-TerminalWraithModel.ps1
    ├── Update-TerminalWraithProvider.ps1
    ├── assets/
    │   ├── terminal_wraith_ascii.txt
    │   ├── terminal_wraith_ascii_plain.txt
    │   └── terminal_wraith_ascii_logo.png
    ├── agent/
    ├── agentics/
    ├── logic/
    ├── pl/
    ├── reasoning/
    ├── research/
    ├── scratch/
    └── security/

Module roles

  • Launcher - starts Terminal Wraith, prints the ASCII banner, and exposes the menu.
  • Recursive Daemon / Agent - scans projects, proposes interventions, applies patches, runs tests, and records evidence.
  • Reasoning - supports local reasoning, verification summaries, causal analysis, and experiment logs.
  • Security - checks for secrets, dependency issues, insecure coding patterns, and threat-model concerns.
  • Program Logic - provides syntax checks, static checks, testing, symbolic checks, contract checks, and verification reports.
  • Logic - supports proof-state tracking, proof obligations, induction, termination notes, and trust ledgers.
  • Agentics - stores operational trace, project memory, long-horizon settings, and search configuration.
  • Research - records claims, benchmarks, reproducibility notes, counterexamples, and evaluation reports.
  • Assets - contains the Terminal Wraith ASCII banner and logo assets.

Basic workflow

install -> run twraith -> choose mode -> inspect output -> verify evidence -> revise or accept
Download terminal-wraith ai
Download terminal-wraith ai