fix: set GIT_PAGER=cat to prevent git pager CPU spin (fixes #1697)

[RemyLoveLogicAI]

Summary

When Agent Zero's code_execution_tool runs git commands that produce long output (e.g., git diff, git branch -a, git log), git pipes the output through the default pager (more or less). In the non-interactive terminal sessions created by code_execution_tool, the pager waits for user input (space/q) that never comes. The pager then spins at 100% CPU per process indefinitely, consuming cores until the container is restarted.

Fix

In plugins/_code_execution/helpers/shell_local.py, the connect() method now passes GIT_PAGER=cat and PAGER=cat in the environment variables to the TTYSession constructor. This tells git to skip the interactive pager and output directly to stdout.

Changes

• Added import os to shell_local.py
• In connect(): copy the current environment and set GIT_PAGER=cat and PAGER=cat before passing to TTYSession

Fixes #1697

[chatgpt-codex-connector]

💡 Codex Review

https://github.com/agent0ai/agent-zero/blob/cc982cb1d3292ab26ea9b16f40120802f28c13e4/plugins/_code_execution/helpers/shell_local.py#L1
Set the pager in python.helpers.shell_local

This adds an empty helper under plugins/_code_execution, but code_execution_tool imports LocalInteractiveSession from python.helpers.shell_local (python/tools/code_execution_tool.py:8), and repo-wide search finds no imports of this new path. Because the real python/helpers/shell_local.py still starts bash without GIT_PAGER=cat or PAGER=cat, git commands run through local code_execution_tool sessions still hit the default pager and the CPU-spin issue remains unresolved.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Wire the hacking image to the hacking prompts

This states that the hacking image sets AGENT_PROMPTS_SUBDIR=hacking, but the Kali Dockerfile and /exe/run_A0.sh do not set that environment variable or pass --agent_prompts_subdir=hacking, and the settings code defaults agent_prompts_subdir to default. As a result, users who run the documented frdel/agent-zero-run:hacking image still get the default prompt set unless they manually change settings, so the added prompts/hacking/ override is not enabled by the image as documented.

Useful? React with 👍 / 👎.

[RemyLoveLogicAI]

Working on resolving merge conflicts now.