coding-agent

Name: coding-agent
Author: SafeAI-Lab-X

1.0kstars58forks

How to install coding-agent?

safeai-lab-x/clawkeeper/coding-agent

$npx -y skills add SafeAI-Lab-X/ClawKeeper --skill coding-agent

Installs into the current project.

Summary

Delegate coding tasks to Codex, Claude Code, or Pi agents via background process. Use when: (1) building/creating new features or apps, (2) reviewing PRs (spawn in temp dir), (3) refactoring large codebases, (4) iterative coding that needs file exploration. NOT for: simple one-li

Files · 1

View on GitHub

SKILL.md

1	# Coding Agent (bash-first)
2
3	Use bash (with optional background mode) for all coding agent work. Simple and effective.
4
5	## ⚠️ PTY Mode: Codex/Pi/OpenCode yes, Claude Code no
6
7	For Codex, Pi, and OpenCode, PTY is still required (interactive terminal apps):
8
9	```bash
10	# ✅ Correct for Codex/Pi/OpenCode
11	bash pty:true command:"codex exec 'Your prompt'"
12	```
13
14	For Claude Code (`claude` CLI), use `--print --permission-mode bypassPermissions` instead.
15	`--dangerously-skip-permissions` with PTY can exit after the confirmation dialog.
16	`--print` mode keeps full tool access and avoids interactive confirmation:
17
18	```bash
19	# ✅ Correct for Claude Code (no PTY needed)
20	cd /path/to/project && claude --permission-mode bypassPermissions --print 'Your task'
21
22	# For background execution: use background:true on the exec tool
23
24	# ❌ Wrong for Claude Code
25	bash pty:true command:"claude --dangerously-skip-permissions 'task'"
26	```
27
28	### Bash Tool Parameters
29
30	\| Parameter \| Type \| Description \|
31	\| ------------ \| ------- \| --------------------------------------------------------------------------- \|
32	\| `command` \| string \| The shell command to run \|
33	\| `pty` \| boolean \| Use for coding agents! Allocates a pseudo-terminal for interactive CLIs \|
34	\| `workdir` \| string \| Working directory (agent sees only this folder's context) \|
35	\| `background` \| boolean \| Run in background, returns sessionId for monitoring \|
36	\| `timeout` \| number \| Timeout in seconds (kills process on expiry) \|
37	\| `elevated` \| boolean \| Run on host instead of sandbox (if allowed) \|
38
39	### Process Tool Actions (for background sessions)
40
41	\| Action \| Description \|
42	\| ----------- \| ---------------------------------------------------- \|
43	\| `list` \| List all running/recent sessions \|
44	\| `poll` \| Check if session is still running \|
45	\| `log` \| Get session output (with optional offset/limit) \|
46	\| `write` \| Send raw data to stdin \|
47	\| `submit` \| Send data + newline (like typing and pressing Enter) \|
48	\| `send-keys` \| Send key tokens or hex bytes \|
49	\| `paste` \| Paste text (with optional bracketed mode) \|
50	\| `kill` \| Terminate the session \|
51
52	---
53
54	## Quick Start: One-Shot Tasks
55
56	For quick prompts/chats, create a temp git repo and run:
57
58	```bash
59	# Quick chat (Codex needs a git repo!)
60	SCRATCH=$(mktemp -d) && cd $SCRATCH && git init && codex exec "Your prompt here"
61
62	# Or in a real project - with PTY!
63	bash pty:true workdir:~/Projects/myproject command:"codex exec 'Add error handling to the API calls'"
64	```
65
66	Why git init? Codex refuses to run outside a trusted git directory. Creating a temp repo solves this for scratch work.
67
68	---
69
70	## The Pattern: workdir + background + pty
71
72	For longer tasks, use background mode with PTY:
73
74	```bash
75	# Start agent in target directory (with PTY!)
76	bash pty:true workdir:~/project background:true command:"codex exec --full-auto 'Build a snake game'"
77	# Returns sessionId for tracking
78
79	# Monitor progress
80	process action:log sessionId:XXX
81
82	# Check if done
83	process action:poll sessionId:XXX
84
85	# Send input (if agent asks a question)
86	process action:write sessionId:XXX data:"y"
87
88	# Submit with Enter (like typing "yes" and pressing Enter)
89	process action:submit sessionId:XXX data:"yes"
90
91	# Kill if needed
92	process action:kill sessionId:XXX
93	```
94
95	Why workdir matters: Agent wakes up in a focused directory, doesn't wander off reading unrelated files (like your soul.md 😅).
96
97	---
98
99	## Codex CLI
100
101	Model: `gpt-5.2-codex` is the default (set in ~/.codex/config.toml)
102
103	### Flags
104
105	\| Flag \| Effect \|
106	\| --------------- \| -------------------------------------------------- \|
107	\| `exec "prompt"` \| One-shot execution, exits when done \|
108	\| `--full-auto` \| Sandboxed but auto-approves in worksp

People who liked this also viewed

6 picks

Type

coding-agent

1.0kstars58forks

View on GitHub

How to install coding-agent?

safeai-lab-x/clawkeeper/coding-agent

$npx -y skills add SafeAI-Lab-X/ClawKeeper --skill coding-agent

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Coding Agent (bash-first)
2
3	Use bash (with optional background mode) for all coding agent work. Simple and effective.
4
5	## ⚠️ PTY Mode: Codex/Pi/OpenCode yes, Claude Code no
6
7	For Codex, Pi, and OpenCode, PTY is still required (interactive terminal apps):
8
9	```bash
10	# ✅ Correct for Codex/Pi/OpenCode
11	bash pty:true command:"codex exec 'Your prompt'"
12	```
13
14	For Claude Code (`claude` CLI), use `--print --permission-mode bypassPermissions` instead.
15	`--dangerously-skip-permissions` with PTY can exit after the confirmation dialog.
16	`--print` mode keeps full tool access and avoids interactive confirmation:
17
18	```bash
19	# ✅ Correct for Claude Code (no PTY needed)
20	cd /path/to/project && claude --permission-mode bypassPermissions --print 'Your task'
21
22	# For background execution: use background:true on the exec tool
23
24	# ❌ Wrong for Claude Code
25	bash pty:true command:"claude --dangerously-skip-permissions 'task'"
26	```
27
28	### Bash Tool Parameters
29
30	\| Parameter \| Type \| Description \|
31	\| ------------ \| ------- \| --------------------------------------------------------------------------- \|
32	\| `command` \| string \| The shell command to run \|
33	\| `pty` \| boolean \| Use for coding agents! Allocates a pseudo-terminal for interactive CLIs \|
34	\| `workdir` \| string \| Working directory (agent sees only this folder's context) \|
35	\| `background` \| boolean \| Run in background, returns sessionId for monitoring \|
36	\| `timeout` \| number \| Timeout in seconds (kills process on expiry) \|
37	\| `elevated` \| boolean \| Run on host instead of sandbox (if allowed) \|
38
39	### Process Tool Actions (for background sessions)
40
41	\| Action \| Description \|
42	\| ----------- \| ---------------------------------------------------- \|
43	\| `list` \| List all running/recent sessions \|
44	\| `poll` \| Check if session is still running \|
45	\| `log` \| Get session output (with optional offset/limit) \|
46	\| `write` \| Send raw data to stdin \|
47	\| `submit` \| Send data + newline (like typing and pressing Enter) \|
48	\| `send-keys` \| Send key tokens or hex bytes \|
49	\| `paste` \| Paste text (with optional bracketed mode) \|
50	\| `kill` \| Terminate the session \|
51
52	---
53
54	## Quick Start: One-Shot Tasks
55
56	For quick prompts/chats, create a temp git repo and run:
57
58	```bash
59	# Quick chat (Codex needs a git repo!)
60	SCRATCH=$(mktemp -d) && cd $SCRATCH && git init && codex exec "Your prompt here"
61
62	# Or in a real project - with PTY!
63	bash pty:true workdir:~/Projects/myproject command:"codex exec 'Add error handling to the API calls'"
64	```
65
66	Why git init? Codex refuses to run outside a trusted git directory. Creating a temp repo solves this for scratch work.
67
68	---
69
70	## The Pattern: workdir + background + pty
71
72	For longer tasks, use background mode with PTY:
73
74	```bash
75	# Start agent in target directory (with PTY!)
76	bash pty:true workdir:~/project background:true command:"codex exec --full-auto 'Build a snake game'"
77	# Returns sessionId for tracking
78
79	# Monitor progress
80	process action:log sessionId:XXX
81
82	# Check if done
83	process action:poll sessionId:XXX
84
85	# Send input (if agent asks a question)
86	process action:write sessionId:XXX data:"y"
87
88	# Submit with Enter (like typing "yes" and pressing Enter)
89	process action:submit sessionId:XXX data:"yes"
90
91	# Kill if needed
92	process action:kill sessionId:XXX
93	```
94
95	Why workdir matters: Agent wakes up in a focused directory, doesn't wander off reading unrelated files (like your soul.md 😅).
96
97	---
98
99	## Codex CLI
100
101	Model: `gpt-5.2-codex` is the default (set in ~/.codex/config.toml)
102
103	### Flags
104
105	\| Flag \| Effect \|
106	\| --------------- \| -------------------------------------------------- \|
107	\| `exec "prompt"` \| One-shot execution, exits when done \|
108	\| `--full-auto` \| Sandboxed but auto-approves in worksp

People who liked this also viewed

6 picks

Type