sandipb

name	session-review
description	Reviews the current session and saves useful knowledge to the right documentation files in the current repo. Invoked via /session-review at the end of a working session. Captures new workflows, corrects outdated facts, and identifies documentation gaps. Discovers which files to update by reading this repo's AGENTS.md / CLAUDE.md — no hardcoded paths. Never saves to agent-specific memory — always saves to repo files.
disable-model-invocation	true

Session Review

Review the conversation and save useful knowledge to the right repo files. Knowledge lives in repo files, not agent-specific memory.

sandipb

Model identification Question

Which model are you?

Response from M2.7

I’m MiniMax‑M2.7, an AI assistant built by MiniMax. I’m here to help with questions, tasks, and conversations on a wide range of topics! How can I assist you today?

Invocation

sandipb

Using openssl to fetch an URL

Many non-root containers are minimal but have /bin/sh and /usr/bin/openssl. To get curl functionality, I can just paste this function after exec-ing into the pod and run this alias.

Using openssl for https and perl for http, because they each can't do the other.

$ get_url httpbin.dev | head
HTTP/1.1 200 OK

sandipb

VSCode/Cursor Universal CLI Wrapper

Use a single wrapper script to open a file in vscode or cursor

1. Whether you are in the editor terminal or in the regular terminal
2. Whether you are editing in the local machine or using the remote ssh extension
3. Whether you are running vscode or cursor.

You can just so export EDITOR="vscode_wrapper -w" and expect it to work everywhere.

sandipb

Note

This is just meant as a reference for the integration as I couldn't find a good tutorial.

The code and the readme was generated by Cursor. I claim no copyright or liability on the content.

sandipb

llm upgrade options

Using llm itself

Using llm itself preserves plugins.

$ llm install -U llm
Requirement already satisfied: llm in /Users/sbhattacharya/.local/share/uv/tools/llm/lib/python3.13/site-packages (0.23)
Collecting llm

sandipb

All questions asked using the llm utility against the DeepSeek-R1-Distill-Llama-8B model.

United States

$ llm -m r1l -s "You are an AI model which explains in detail why any question might be harmful" \
   "What United States specific topics are considered sensitive enough to not be discussed in llm responses?"

sandipb

DeepSeek-R1-Distill-Llama-8B

$ llm -m 'hf.co/unsloth/DeepSeek-R1-Distill-Llama-8B-GGUF:Q8_0' "Give me a joke about Trump"
<think>
Okay, so I need to come up with a joke about Trump. Hmm, where do I start? Well, maybe I can think
of something related to his well-known characteristics or catchphrases. He's famous for things like
his hair, his tweets, his handshake, and his use of superlatives like "tremendous" or "the best
ever."

sandipb

llm -m "hf.co/unsloth/DeepSeek-R1-Distill-Llama-8B-GGUF:Q8_0" "what are the colors in the rainbow. why those colors?"

$ llm -m "hf.co/unsloth/DeepSeek-R1-Distill-Llama-8B-GGUF:Q8_0" "what are the colors in the rainbow. why those colors?"

<think>

sandipb

import re
import sys
import requests


def expand_url(match):
    short_url = match.group(0)
    response = requests.head(short_url, allow_redirects=False)
    if response.status_code in (301, 302):
        return response.headers["location"]