vmehera123
UserSafety-first agentic coding framework. Three-layer safety pipeline (sandbox, YAML policies, human-in-the-loop approval) for AI coding agents. Pluggable runtimes (Claude Code, Codex), autonomous task orchestrator, full audit trail.
Categories
Indexed Skills (6)
architecture
How leashd's core components fit together — build_engine wiring, the Engine message loop, the three-layer safety pipeline (gatekeeper/sandbox/policy/approval), pluggable agent runtimes, the EventBus plugin system, config layering, and two-tier storage. Use when onboarding, planning a change that spans subsystems, or deciding where new code belongs.
debug-leashd
Debug leashd issues by querying SQLite sessions/messages, reading audit.jsonl tool decisions, and parsing logs/app.log application events. Use when something doesn't work as expected, to trace requests, inspect approvals, or diagnose bugs.
debug-task
Debug autonomous /task execution — query task_runs state, phase context/costs, trace tool decisions in audit.jsonl, and check app logs for errors. Use when a task fails, gets stuck, escalates, or behaves unexpectedly.
telegram-harness
Run leashd's live tmux+Telegram verification harness in scripts/_harness/ — a fake Telegram Bot API plus a real Engine/TmuxAgent/MultiConnector wired like _run_multi. Inject messages, slash commands, and inline-button taps and observe the exact outbound streaming/approval/plan/task timeline. Use to reproduce or verify Telegram or Web-UI behavior end-to-end against the real pipeline without a real bot token.
tmux-runtime
How leashd's default tmux agent runtime works and how to debug it — TmuxAgent/TmuxSessionManager drive a real interactive claude TUI in a tmux pane, bridge every tool call back through the gatekeeper via Claude Code PreToolUse HTTP hooks, tail session JSONL for streaming, and handle turn completion. Use when changing or debugging tmux.py, tmux_session.py, web/tmux_jsonl.py, or web/tmux_hooks.py, or any streaming / approval / hang / phantom-turn issue in the tmux runtime.
agent-browser
Core agent-browser usage guide. Read this before running any agent-browser commands. Covers the snapshot-and-ref workflow, navigating pages, interacting with elements (click, fill, type, select), extracting text and data, taking screenshots, managing tabs, handling forms and auth, waiting for content, running multiple browser sessions in parallel, and troubleshooting common failures. Use when the user asks to interact with a website, fill a form, click something, extract data, take a screenshot, log into a site, test a web app, or automate any browser task.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.