Shell Gpt

Run AI-powered shell commands and prompts for faster terminal workflows. Use when generating scripts, comparing outputs, evaluating prompts.

Security Scan

The skill claims to be an AI-powered shell tool but the included code only implements local logging and does not integrate with any LLMs; it also encourages storing API keys/prompts in plaintext logs, which is inconsistent and risky.

Purpose & Capability

The name and SKILL.md claim an "AI-powered" CLI for running and evaluating model prompts and mention GPT-5/model benchmarking and storing API keys, but the provided script implements only local logging/management of text entries (writes logs under ~/.local/share/shell-gpt) and contains no network calls or LLM integration. The skill's requested artifacts (none) do not match the described capability to call or configure LLMs.

Instruction Scope

Runtime instructions and examples encourage persisting prompts, API keys, and other sensitive entries via the `configure` command, but the script simply appends those inputs to plaintext log files. The instructions give broad permission to store arbitrary content (including secrets) without guidance on encryption or access controls — this is scope creep relative to a benign local journaling tool and could lead to secret leakage.

Install Mechanism

No install spec; the skill is instruction-only with a single shell script. There are no downloads or third-party package installs. The script will create and write files under the user's home directory, which is expected for a local CLI tool.

Credentials

The skill declares no required environment variables or credentials, yet SKILL.md explicitly suggests persisting API keys and model preferences via `configure`. That mismatch is concerning: the skill does not request or protect credentials but encourages storing them in plaintext logs, which is disproportionate and risky.

Persistence & Privilege

always is false and the skill is user-invocable only; it does not request elevated or cross-skill privileges. It does, however, persist data under $HOME/.local/share/shell-gpt — reasonable for a CLI but worth noting because sensitive data may be stored there without encryption.

What to consider before installing

This package looks internally inconsistent: its description promises an "AI-powered" assistant that interacts with LLMs, but the included script only provides a local logging/journal CLI and does not call any external models. The SKILL.md suggests saving API keys and prompts via `configure`, but the script appends those inputs to plaintext log files under ~/.local/share/shell-gpt. Before installing or using this skill, consider the following: - Do not paste API keys, tokens, or other secrets into the `configure` command or any prompts unless you accept they will be stored in plaintext in your home directory. Prefer storing keys in environment variables or a secure secret manager. - Inspect the script yourself (or run it in a sandbox) to confirm it matches your expectations; the provided script is short and readable and currently has no network calls, but that also means it is not actually contacting any LLM service as the README claims. - If you expected an LLM-integrated tool, ask the publisher for clarification or a version that documents the actual integration and the exact environment variables it requires (and how credentials are protected). If the developer intends to add network/LLM calls, request transparency about endpoints, required env vars, and how secrets are stored. Given the mismatch between claims and implementation and the risk of plaintext secret storage, treat this skill with caution. If you proceed, avoid entering any sensitive credentials into the tool.

A command-line productivity tool powered by AI large language models like GPT-5, designed to help you accomplish terminal tasks faster and more efficiently. Log, track, and manage AI-related workflows including prompt evaluation, model benchmarking, cost tracking, and output analysis — all from a single CLI.

Inspired by TheR1D/shell_gpt (11,891+ GitHub stars).

Commands

• configure — Set up or review configuration parameters for Shell Gpt. Run without arguments to view recent config entries, or pass key-value pairs to save new settings. Useful for persisting API keys, model preferences, or runtime flags.

• benchmark — Record and review benchmark results for model or prompt performance. Without arguments it shows recent benchmark logs; with arguments it saves a new benchmark entry with a timestamp. Great for tracking latency, throughput, or quality metrics over time.

• compare — Log side-by-side comparison notes between models, prompts, or outputs. View previous comparisons with no arguments, or save a new comparison record. Helps you keep a structured journal of A/B test results.

• prompt — Store and retrieve prompt templates or prompt-engineering notes. Run bare to see recent prompts, or pass text to save a new prompt entry. Essential for iterating on prompt design and keeping a prompt library.

• evaluate — Record evaluation scores or qualitative assessments of model outputs. Without arguments it displays recent evaluations; with arguments it logs a new evaluation. Use this to build a track record of output quality over time.

• fine-tune — Track fine-tuning jobs, hyperparameters, and results. Shows recent fine-tune logs when called alone, or saves a new entry when given input. Handy for documenting training runs alongside their outcomes.

• analyze — Log analysis observations about data, outputs, or model behavior. View past analyses or record a new one. Use this as a lightweight lab notebook for your AI experiments.

• cost — Track API costs, token usage expenses, and billing notes. Without arguments it shows recent cost entries; with input it saves a new cost record. Critical for keeping cloud AI spending under control.

• usage — Record and review API usage statistics such as token counts or request volumes. Displays recent usage entries when called bare, or logs new usage data when given arguments. Helps you stay within rate limits and quotas.

• optimize — Document optimization attempts, parameter tweaks, and their effects. Shows recent optimization logs or saves a new entry. Use this to track what you tried and what worked.

• test — Log test cases, test results, and validation outcomes. View recent test entries or record a new one. Perfect for building a regression test journal for your prompts or pipelines.

• report — Save and review summary reports or periodic write-ups. Without arguments it shows recent reports; with text it saves a new report entry. Ideal for weekly summaries or stakeholder updates.

• stats — Display summary statistics across all logged categories. Shows total entry counts per category, overall totals, and data directory disk usage. A quick dashboard for your Shell Gpt activity.

• export <fmt> — Export all logged data to a file in the specified format: json, csv, or txt. The exported file is saved to the data directory and its size is reported. Use this for backups, data migration, or feeding logs into other tools.

• search <term> — Search across all log files for a given term (case-insensitive). Returns matching lines grouped by category. Helpful for finding a specific entry when you remember a keyword but not which command logged it.

• recent — Show the 20 most recent entries from the global activity history log. Gives you a quick at-a-glance view of what has been happening across all commands.

• status — Run a health check on the Shell Gpt installation. Reports the current version, data directory path, total log entries, disk usage, last activity timestamp, and overall status. Use this to verify the tool is working correctly.

• help — Display the full usage guide listing all available commands and their syntax. Also shows the current data directory path.

• version — Print the current Shell Gpt version string (shell-gpt v2.0.0).

When to Use

• Prompt engineering sessions — When you are iterating on prompt designs, use prompt to save drafts, evaluate to score outputs, and compare to log A/B differences between prompt variants.
• Model benchmarking and selection — When comparing multiple LLMs or model versions, use benchmark to record performance metrics, analyze to note behavioral observations, and report to write up your conclusions.
• Cost and usage monitoring — When you need to keep cloud AI spending in check, use cost to log expenses, usage to track token consumption, and stats to get an overview of activity volume.
• Fine-tuning workflow tracking — When running fine-tuning jobs, use fine-tune to document hyperparameters and results, test to record validation outcomes, and optimize to log parameter adjustments.
• Data export and auditing — When you need to share logs, create backups, or feed data into dashboards, use export to dump everything to JSON/CSV/TXT, and search to locate specific entries quickly.

Examples

# Save a new prompt template and review recent prompts
shell-gpt prompt "Summarize the following text in 3 bullet points:"
shell-gpt prompt

# Benchmark a model and check overall stats
shell-gpt benchmark "gpt-4o: 320ms avg latency, 92% accuracy on eval set"
shell-gpt stats

# Log API costs and export all data as CSV
shell-gpt cost "2026-03-18: $4.27 for 1.2M tokens on gpt-4o"
shell-gpt export csv

# Compare two models side by side
shell-gpt compare "gpt-4o vs claude-opus: gpt-4o faster, claude-opus more detailed"

# Search across all logs for a keyword and check system health
shell-gpt search "latency"
shell-gpt status

Output

Returns reports and log entries to stdout. All data is persisted to per-command log files in the data directory. Redirect output to a file with shell-gpt <command> > output.txt.

Configuration

Set the SHELL_GPT_DIR environment variable to change the data directory. Default: ~/.local/share/shell-gpt/. All log files and exports are stored in this directory.

---

Powered by BytesAgain | bytesagain.com Feedback & Feature Requests: https://bytesagain.com/feedback