All skills
simota avatar

/hone

@c805268
by shingo imotasimota/agent-skills85 stars
15

Auditing AI CLI configs and designing, configuring, or debugging Claude Code hooks. Use for Codex/agy/Claude Code config reviews, hook lifecycle automation, quality gates, or MCP governance.

Use this Skill: https://skilld.dev/gh/simota/agent-skills/hone

This session only. Nothing lands on disk.

referencecodex-config-schema.md

≈1.9k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Codex Config Schema

Purpose: Comprehensive reference for all Codex CLI configuration keys, default values, and recommended settings. Read when: Auditing ~/.codex/config.toml or proposing configuration changes. Last updated: 2026-03-12 (Codex CLI v0.114.0)


Config File Location

Primary: ~/.codex/config.toml


Core Settings

Key Type Default Description Recommendation
model string Account/runtime dependent Supported authorized model ID Verify current stable availability in _common/CLI_COMPATIBILITY.md; inherit explicit user selection
provider string "openai" API provider Match to model availability
model_reasoning_effort string Model/runtime dependent Supported reasoning effort Verify current official schema and effective settings; no automatic maximum or spend escalation
model_reasoning_summary string "auto" Reasoning summary: auto, concise, detailed, none concise for transparency without verbosity
model_verbosity string "medium" Output verbosity: low, medium, high medium for balanced output
personality string "none" Communication style: none, friendly, pragmatic pragmatic for professional use
hide_agent_reasoning boolean false Suppress reasoning events Keep false for transparency
show_raw_agent_reasoning boolean false Runtime-specific diagnostic setting Do not enable by default or request private reasoning reproduction; use observable decisions and tool evidence

Policy Settings

Key Type Default Description Recommendation
approval_policy string "on-request" Approval mode: on-request, never, unless-allow-listed on-request for safety
sandbox_mode string "workspace-write" Sandbox: workspace-write, off, read-only workspace-write for balanced security

TUI Settings

[tui]
status_line = ["model-with-reasoning", "context-remaining", "current-dir"]
theme = "catppuccin-mocha"
animations = true
show_tooltips = true
alternate_screen = "auto"
notifications = true
notification_method = "auto"

Status Line Items (verified against v0.114.0)

Item ID Displayed Content Notes
model Model name
model-with-reasoning Model name + reasoning level included by default
model-name Model name (without reasoning)
git-branch Git branch name fetched asynchronously
context-remaining Remaining context included by default
context-used Context used
context-window-size Context window size
current-dir Working directory included by default
project-root Project root name
fast-mode Fast mode state
session Session ID
version Codex version

Note: IDs defined in PR #10546 that may not yet be implemented in v0.114.0: rate-limits, tokens, tokens-input, tokens-output, cwd. Use the /statusline command to check which items are actually available.

  • null or [] hides the status line
  • Can be configured interactively via the /statusline command

TUI Options

Key Type Default Description
tui.status_line array<string> | null ["model-with-reasoning", "context-remaining", "current-dir"] Status line items
tui.theme string auto-detect Theme name (kebab-case). Selectable via /theme
tui.animations boolean true Animations for welcome screen, spinners, etc.
tui.show_tooltips boolean true Onboarding tooltips
tui.alternate_screen string "auto" Alternate screen buffer: auto, always, never
tui.notifications boolean | array true Enable desktop notifications / event filter
tui.notification_method string "auto" Notification method: auto, osc9, bel

Custom Themes

Place .tmTheme files in $CODEX_HOME/themes/ to make them selectable via /theme.

Trust Levels

[projects."/path/to/project"]
trust_level = "trusted"      # full-auto allowed
Level Behavior Use when
"trusted" Allows full-auto mode Personal projects, well-tested repos
"untrusted" Forces suggest mode Third-party code, sensitive repos

Trust Level Audit Criteria

  • Stale paths: Projects that no longer exist on disk
  • Over-trust: Sensitive dirs (home, .ssh/, .aws/) marked as trusted
  • Under-trust: Personal projects unnecessarily restricted
  • Wildcard paths: Overly broad trust patterns
  • External volumes: Removable/network paths that may not be mounted

Feature Flags

[features]
unified_exec = true
shell_snapshot = true
steer = true
apps = true
personality = true
multi_agent = true
prevent_idle_sleep = true

Feature flags evolve rapidly. Verify against the latest release notes.

History

[history]
persistence = "save-all"    # "save-all", "save-none", "ask"

MCP Server Configuration

[mcp_servers.server-name]
command = "/absolute/path/to/binary"
args = ["arg1", "arg2"]
env = { KEY = "value" }

MCP Audit Criteria

  • Accessibility: Server binary exists and is executable
  • Path: Use absolute paths to avoid PATH resolution issues
  • Version: Server is up-to-date
  • Necessity: Server is actively used
  • Security: No secrets in plaintext args/env

Rules

# ~/.codex/rules/default.rules
prefix_rule(pattern=["git", "add"], decision="allow")
prefix_rule(pattern=["git", "commit", "-m"], decision="allow")

Rules Audit Criteria

  • Overly specific: Individual command rules that should be generalized
  • Redundant: Rules subsumed by broader patterns
  • Missing: Common commands that lack allow rules

Other Configuration Files

File Purpose Audit focus
~/.codex/instructions.md Global instructions Non-empty, clear, up-to-date
~/.codex/AGENTS.md Agent behavior definitions Priority clarity, no redundancy
~/.codex/rules/ Rule files (glob-scoped) No duplicates, valid patterns
~/.codex/config.json Legacy config (pre-TOML) Should be deleted if present

Other Display Settings

Key Type Default Description
file_opener string "vscode" Editor for file citations: vscode, vscode-insiders, windsurf, cursor, none
disable_paste_burst boolean false Disable burst-paste detection
web_search string — Web search: cached, always, never

Profiles

[profiles.ci]
approval_policy = "never"
sandbox_mode = "workspace-write"

[profiles.review]
model_reasoning_effort = "high"
model_reasoning_summary = "detailed"

Usage: codex --profile ci

Notice (Auto-managed)

[notice]
hide_xxx_migration_prompt = true

[notice.model_migrations]
"old-model" = "new-model"

A section auto-managed by Codex CLI. It's safe to manually remove stale migration prompts.


Version History

Date Change Source
2026-03-12 Full schema update for v0.114.0: added TUI, Policy, History, Profiles, model_reasoning_summary, approval_policy, sandbox_mode T1+T2
Initial Schema documented from Codex CLI v0.1 T2: openai/codex GitHub

This schema should be refreshed when major Codex CLI versions are released.

Source: SKILL.md on GitHub

1 warning13d4 checks · Risk SAFE
  • Gen Agent Trust Hub13d

    The skill is a security-focused configuration auditor and automation tool for AI CLI environments. It includes extensive features for detecting hardcoded secrets, blocking dangerous shell commands, and managing environment security through best practices. No malicious patterns were detected.

  • Socket13d

    No alerts

  • Snyk13d

    Risk: LOW · No issues

  • ZeroLeaks5mo

    1 finding · Score: 78/100

Signed by skilld at c805268. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated 2 weeks ago

README badge

README badge for simota/agent-skills/hone