niahere 0.2.35 → 0.2.36

package/README.md

@@ -1,11 +1,25 @@
 # nia
 
-A personal AI assistant that runs as a background daemon. Handles scheduled jobs, terminal chat, Telegram, and Slack — powered by Claude.
+A personal AI agent you fork and make your own. Small enough to understand, built for one user. Powered by Claude Agent SDK.
 
 - npm package: [`niahere`](https://www.npmjs.com/package/niahere)
 - CLI command: `nia`
 - Website: [niahere.com](https://niahere.com)
 
+## Philosophy
+
+**Small enough to understand.** One process, a few source files. No microservices, no message queues, no abstraction layers. Have Claude Code walk you through it.
+
+**Built for one user.** This isn't a framework. It's working software that fits your exact needs. You fork it and have Claude Code make it match your exact needs.
+
+**Customization = code changes.** No configuration sprawl. Want different behavior? Modify the code. The codebase is small enough that this is safe.
+
+**AI-native.** No installation wizard; Claude Code guides setup. No monitoring dashboard; ask Claude what's happening. No debugging tools; describe the problem, Claude fixes it.
+
+**Skills over features.** Contributors shouldn't add features to the codebase. Instead, they contribute claude code skills like `/add-discord` that transform your fork. You end up with clean code that does exactly what you need.
+
+**Best harness, best model.** This runs on Claude Agent SDK, which means you're running Claude Code directly. The harness matters. A bad harness makes even smart models seem dumb, a good harness gives them superpowers.
+
 ## Quick Start
 
 ```bash
@@ -14,6 +28,18 @@ nia init                # guided setup (database, channels, persona, visual iden
 nia start               # starts daemon + registers OS service
 ```
 
+## What It Supports
+
+- **Telegram** — message your agent from your phone, typing indicator while processing
+- **Slack** — Socket Mode bot with thread awareness, thinking emoji, watch channels for proactive monitoring
+- **Terminal chat** — REPL with session resume support
+- **Scheduled jobs** — recurring jobs and crons that run Claude and can message you back
+- **Persona system** — customizable identity, soul, owner profile, and on-demand memory
+- **Skills** — loads skills from multiple directories, invokable as slash commands
+- **Cross-platform service** — launchd (macOS), systemd (Linux), service-aware restart
+- **MCP tools** — 18 tools for job management, messaging, memory, and channel control
+- **Optional integrations** — add Gmail, Discord, and more via skills
+
 ## Commands
 
 ```
@@ -50,30 +76,6 @@ nia channels                   — show channel status (on/off)
 nia channels on / off          — enable/disable channels
 ```
 
-## Features
-
-- **Jobs & crons** — jobs run during active hours, crons run 24/7. Stored in PostgreSQL, auto-reload via LISTEN/NOTIFY. One-shot jobs auto-disable after execution. Full JSONL traces with Codex session IDs.
-- **Terminal chat** — REPL with session resume support
-- **Telegram** — bot with access control, typing indicator while processing
-- **Slack** — Socket Mode bot with thinking emoji reactions, thread awareness (auto-listens to follow-ups without @mention), thread context fetching, owner vs non-owner access control, prompt injection defense
-- **Persona system** — customizable identity, soul, owner profile, and on-demand memory
-- **Visual identity** — AI-generated profile pictures via Gemini, customizable during `nia init`
-- **Cross-platform service** — launchd (macOS), systemd (Linux), service-aware restart
-- **Skills** — loads skills from `~/.shared/skills/`, `~/.claude/skills/`, `~/.codex/skills/`, and bundled skills
-- **Dev mode** — `nia channels off` disables Telegram/Slack for local development without conflicts
-
-## Updating
-
-```bash
-npm i -g niahere         # pulls the latest version from npm
-```
-
-To publish a new version after making changes:
-
-```bash
-npm run release          # bumps patch version, publishes to npm, pushes git tag
-```
-
 ## Architecture
 
 All config and data lives in `~/.niahere/`:
@@ -93,6 +95,14 @@ All config and data lives in `~/.niahere/`:
     nia.pid, daemon.log, cron-state.json, cron-audit.jsonl
 ```
 
+## Contributing
+
+**Don't add features. Add skills.**
+
+If you want to add Discord support, don't create a PR that adds Discord alongside Telegram. Instead, contribute a skill folder (`skills/add-discord/SKILL.md`) that teaches Claude Code how to transform a nia installation to use Discord.
+
+Users then run `/add-discord` on their fork and get clean code that does exactly what they need, not a bloated system trying to support every use case.
+
 ## Requirements
 
 - [Bun](https://bun.sh) runtime (auto-installed if missing)
@@ -101,6 +111,12 @@ All config and data lives in `~/.niahere/`:
 - Gemini API key (optional, for image generation — `nia config set gemini_api_key ...`)
 - OpenAI API key (optional, for image generation — `nia config set openai_api_key ...`)
 
+## Updating
+
+```bash
+npm i -g niahere         # pulls the latest version from npm
+```
+
 ## Author
 
 Aman ([amankumar.ai](https://amankumar.ai))

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "niahere",
-  "version": "0.2.35",
+  "version": "0.2.36",
   "description": "A personal AI assistant daemon — scheduled jobs, chat across Telegram and Slack, persona system, and visual identity.",
   "type": "module",
   "scripts": {

package/skills/taskmaster/SKILL.md

@@ -0,0 +1,104 @@
+---
+name: taskmaster
+description: |
+  Completion guard that prevents premature task abandonment. Use when wrapping up
+  any non-trivial task, before claiming work is "done", or when you catch yourself
+  writing a summary instead of finishing the work. Also use when the user says
+  "don't stop early", "finish everything", "make sure it's actually done", or
+  "verify completion". Adapted from github.com/blader/taskmaster.
+---
+
+# Taskmaster
+
+Completion guard. Progress is not completion. Before you stop, prove the work is done.
+
+## When to Invoke
+
+- Before claiming any non-trivial task is complete
+- When you're about to write a summary of what you did
+- When you notice yourself using phrases like "significant progress", "mostly done", or "remaining work would require..."
+- When the user explicitly asks you to finish everything
+
+## The Completion Checklist
+
+Run every item. Do not skip any. Do not summarize — execute.
+
+### 1. Goal Confrontation (Do This First)
+
+Answer these three questions explicitly in your response:
+
+a. **What is the user's stated goal or success criterion?** Write it out verbatim.
+b. **Is it achieved RIGHT NOW?** Answer "yes" or "no". Not "partially", not "mostly", not "significant progress was made". Yes or no.
+c. **If no:** you are NOT DONE. Go do more work.
+
+The ONLY exception is if the user explicitly told you to stop or deprioritized the goal. There is no other valid reason to stop.
+
+### 2. Re-read the Original Request
+
+Go back to the user's original message(s). List every discrete request or acceptance criterion. For each one, confirm it is **fully addressed** — not started, not in progress, FULLY done.
+
+If the user changed their mind or withdrew a request, treat it as resolved.
+
+### 3. Check the Task List
+
+Review every task. Any task not marked completed? Do it now — unless the user said to skip it.
+
+### 4. Check the Plan
+
+Walk through each step of the plan, INCLUDING verification steps. Any step skipped or partially done? Finish it.
+
+If the plan includes verification steps (builds, tests, lints, type-checks, smoke tests), you MUST actually run them and see them pass. Do not skip them or claim they pass without evidence.
+
+### 5. Check for Errors
+
+Did anything fail or remain unfinished? Fix it. This applies to ALL types of problems — logic errors, missing functionality, incomplete refactors, broken scripts, configuration issues, or anything else that prevents the work from being fully done.
+
+### 6. Check for Loose Ends
+
+- TODO comments you left behind?
+- Placeholder code?
+- Missing tests for new code?
+- Untested changes?
+- Follow-ups you noted but didn't act on?
+
+If any exist, resolve them now.
+
+### 7. Check for Blockers
+
+If something is blocking you, do NOT give up. Try a different approach, read more code, search for examples, re-examine your assumptions.
+
+"I didn't cause this bug" is not an excuse to stop — if it blocks your task, fix it. You own the outcome, not just your diff.
+
+## Anti-Rationalization Guide
+
+These are NOT valid reasons to stop:
+
+| Rationalization | Reality |
+|---|---|
+| "Diminishing returns" | The goal isn't met. Keep working. |
+| "Significant progress was made" | Progress is not completion. |
+| "Would require broader architectural changes" | Then make them. |
+| "No single dominant hotspot" | Keep looking. |
+| "Tried N approaches" | Try N+1. |
+| "I can't do X" | You haven't tried X yet. |
+| "This is a pre-existing issue" | If it blocks your task, it's your issue. |
+| "The remaining work is complex" | That's not a reason to stop. That's a description of your job. |
+
+## Do Not Narrate — Execute
+
+If any incomplete work remains, your ONLY job is to DO that work right now.
+
+- Do NOT respond by explaining what the remaining tasks are
+- Do NOT describe their complexity or list their dependencies
+- Do NOT ask for permission to proceed
+- Do NOT write summaries of what is left
+
+Open files, write code, run commands, fix bugs. Act.
+
+## User Instructions Override
+
+The user's latest instructions always take priority. If the user said to stop, move on, or skip something — respect that. Do not force completion of work the user no longer wants.
+
+## Attribution
+
+Core ideas adapted from [Taskmaster](https://github.com/blader/taskmaster) by blader (MIT).

package/src/commands/init.ts

@@ -7,6 +7,7 @@ import { resetConfig } from "../utils/config";
 import { runMigrations } from "../db/migrate";
 import { closeDb } from "../db/connection";
 import { startDaemon, isRunning } from "../core/daemon";
+import { registerService, isServiceInstalled } from "./service";
 import { errMsg } from "../utils/errors";
 import { enrichSlackConfig } from "../cli/channels";
 import yaml from "js-yaml";
@@ -434,12 +435,25 @@ export async function runInit(): Promise<void> {
 
     resetConfig();
 
+    // Install system service (launchd/systemd) for auto-restart on crash
+    if (!isServiceInstalled()) {
+      try {
+        await registerService();
+        console.log(`\n  \u2713 installed system service (auto-restart on crash)`);
+      } catch (err) {
+        console.log(`\n  \u26a0 could not install system service: ${errMsg(err)}`);
+        console.log(`    run 'nia service install' manually for auto-restart`);
+      }
+    } else {
+      console.log(`\n  - system service already installed`);
+    }
+
     // Auto-start daemon
     if (!isRunning()) {
       const pid = startDaemon();
-      console.log(`\n  \u2713 nia started (pid: ${pid})`);
+      console.log(`  \u2713 nia started (pid: ${pid})`);
     } else {
-      console.log(`\n  - nia already running`);
+      console.log(`  - nia already running`);
     }
 
     console.log("\nDone.");

package/src/commands/service.ts

@@ -8,7 +8,15 @@ const PLIST_NAME = "com.niahere.agent";
 const SYSTEMD_UNIT = "niahere.service";
 
 function getExecCommand(): [string, string] {
-  return [process.execPath, process.argv[1]];
+  const execPath = process.execPath;
+  // process.argv[1] may be undefined when called outside the normal CLI flow
+  // (e.g. from bun -e or programmatic import). Fall back to the known entry point.
+  let cliPath = process.argv[1];
+  if (!cliPath || cliPath === "undefined") {
+    const { resolve } = require("path");
+    cliPath = resolve(import.meta.dir, "../cli/index.ts");
+  }
+  return [execPath, cliPath];
 }
 
 // --- macOS launchd ---

package/src/core/daemon.ts

@@ -162,6 +162,20 @@ export async function runDaemon(): Promise<void> {
     }
   }
 
+  // Crash handlers — ensure PID cleanup and logging on unhandled errors.
+  // Without these, an unhandled rejection kills the process silently,
+  // leaving a stale PID file that blocks restarts and hides the cause.
+  process.on("uncaughtException", (err) => {
+    log.fatal({ err }, "uncaught exception — cleaning up");
+    removePid();
+    process.exit(1);
+  });
+  process.on("unhandledRejection", (reason) => {
+    log.fatal({ reason }, "unhandled rejection — cleaning up");
+    removePid();
+    process.exit(1);
+  });
+
   writePid(process.pid);
   log.info({ pid: process.pid }, "daemon started");