📦

codex

Name: codex
Author: softaworks

Content revision r1 High Risk 📁 Filesystem access⚙️ External commands

Run Codex From Claude Code

Claude Code users often need deeper code analysis or automated edits from Codex. This skill gives Claude a structured way to launch Codex, choose safe execution settings, and summarize results.

Supports: Claude Codex Code(CC)

⚠️ 38 Poor

Stop for confirmation before installing.

Review the plan and obtain explicit user consent before changing files.

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request

Review the Skillstore skill "codex" from https://skillstore.io/skills/softaworks-codex.md and its manifest at https://skillstore.io/api/skills/softaworks-codex/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.

Your Agent should still show its plan and request any confirmation required by the security policy.

Agent-readable resources

Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.

Markdown detail GET /skills/softaworks-codex.md Signed manifest GET /api/skills/softaworks-codex/manifest Signed lockfile GET /api/skills/softaworks-codex/lockfile

Test it

Using "codex". Use Codex to analyze this Claude Code skill repository.

Expected outcome:

A concise review with security notes, implementation risks, and recommended follow-up actions.

Using "codex". Resume the last Codex session and continue the refactor review.

Expected outcome:

A continuation summary that preserves the prior session context and lists next decisions.

Using "codex". Use Codex to apply the approved bug fix.

Expected outcome:

A change summary with affected files, verification results, and any unresolved concerns.

Security Audit

High Risk

v6 • 7/9/2026 Open versioned report

Most static hits are Markdown documentation or normal installation paths, but the skill does enable real external Codex execution. The main risks are full-auto broad-access modes, shell prompt piping during resume, and default stderr suppression.

Files scanned

126

Lines analyzed

Review items

False positives ignored

Confirmed security concerns (5)

High

SKILL.md:29

Network reconnaissance

| Permit network or broad access | `danger-full-access` | `--sandbox danger-full-access --full-auto

The skill documents danger-full-access with full-auto for network or broad access. That mode can permit broad filesystem and network actions by an autonomous coding agent.

High

SKILL.md:10 SKILL.md:28-29 SKILL.md:59

High-Impact Autonomous Codex Execution

The skill enables Codex runs with full-auto and danger-full-access for broad filesystem or network access. User approval is required, but the capability is high impact.

Multiple instructions explicitly mention sandbox selection, full-auto, and danger-full-access. The text includes a permission gate, so intent appears operational rather than malicious.

High

SKILL.md:19 SKILL.md:30 SKILL.md:54

Unsafe Shell Prompt Piping Pattern

The resume workflow shows user prompt text being piped through echo into a shell command. Without careful escaping, prompt text could alter shell behavior.

The examples use shell echo with placeholder prompt text. The risk depends on implementation, but agents may interpolate untrusted user prompts directly.

Medium

README.md:30 README.md:51 SKILL.md:20

Default Suppression Of Codex Diagnostics

The skill tells agents to hide stderr from Codex commands by default. This can conceal warnings, failures, or safety-relevant diagnostic output from review.

The documentation repeatedly states that stderr is redirected to /dev/null. It gives a benign rationale, but diagnostics can contain important warnings.

Low

README.md:30

System reconnaissance

By default, this skill suppresses thinking tokens (stderr output) using `2>/dev/null` to avoid bloat

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Capability review items (25)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

High

SKILL.md:14

Ruby/shell backtick execution

- `--sandbox <read-only|workspace-write|danger-full-access>`

The skill documents danger-full-access with full-auto for network or broad access. That mode can permit broad filesystem and network actions by an autonomous coding agent.

High

SKILL.md:19

Ruby/shell backtick execution

4. When continuing a previous session, use `codex exec --skip-git-repo-check resume --last` via stdi

The resume workflow pipes a prompt through a shell command into codex exec. If user text is interpolated unsafely, shell metacharacters could affect execution.

High

SKILL.md:29

Ruby/shell backtick execution

| Permit network or broad access | `danger-full-access` | `--sandbox danger-full-access --full-auto

The skill documents danger-full-access with full-auto for network or broad access. That mode can permit broad filesystem and network actions by an autonomous coding agent.

High

SKILL.md:30

Ruby/shell backtick execution

| Resume recent session | Inherited from original | `echo "prompt" \| codex exec --skip-git-repo-che

The resume workflow pipes a prompt through a shell command into codex exec. If user text is interpolated unsafely, shell metacharacters could affect execution.

High

SKILL.md:54

Ruby/shell backtick execution

- When resuming, pipe the new prompt via stdin: `echo "new prompt" | codex exec resume --last 2>/dev

The resume workflow pipes a prompt through a shell command into codex exec. If user text is interpolated unsafely, shell metacharacters could affect execution.

High

SKILL.md:59

Ruby/shell backtick execution

- Before you use high-impact flags (`--full-auto`, `--sandbox danger-full-access`, `--skip-git-repo-

This line explicitly names high-impact flags and requires permission before use. The permission gate helps, but the skill still enables those risky modes.

Medium

SKILL.md:10

Ruby/shell backtick execution

2. Select the sandbox mode required for the task; default to `--sandbox read-only` unless edits or n

The skill tells the agent to select sandbox modes, including modes needed for edits or network access. This is expected but can materially change system access.

Medium

SKILL.md:15

Ruby/shell backtick execution

- `--full-auto`

The skill includes --full-auto as an approved command flag. This can allow automated changes or command execution with reduced interactive review.

Medium

SKILL.md:17

Ruby/shell backtick execution

- `--skip-git-repo-check`

The skill instructs use of --skip-git-repo-check, including as a default. This weakens a common safety boundary for repository-scoped operations.

Medium

SKILL.md:20

Ruby/shell backtick execution

5. **IMPORTANT**: By default, append `2>/dev/null` to all `codex exec` commands to suppress thinking

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Medium

SKILL.md:27

Ruby/shell backtick execution

| Read-only review or analysis | `read-only` | `--sandbox read-only 2>/dev/null` |

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Medium

SKILL.md:28

Ruby/shell backtick execution

| Apply local edits | `workspace-write` | `--sandbox workspace-write --full-auto 2>/dev/null` |

The skill includes --full-auto as an approved command flag. This can allow automated changes or command execution with reduced interactive review.

Medium

SKILL.md:31

Ruby/shell backtick execution

| Run from another directory | Match task needs | `-C <DIR>` plus other flags `2>/dev/null` |

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Medium

SKILL.md:58

Ruby/shell backtick execution

- Stop and report failures whenever `codex --version` or a `codex exec` command exits non-zero; requ

This line instructs use of the Codex CLI as an external command. The behavior is expected for the skill but remains a real command execution risk.

Medium

SKILL.md:64

Ruby/shell backtick execution

Requires Codex CLI v0.57.0 or later for GPT-5.2 model support. The CLI defaults to `gpt-5.2` on macO

This line instructs use of the Codex CLI as an external command. The behavior is expected for the skill but remains a real command execution risk.

Low

README.md:30

Standard device file access

By default, this skill suppresses thinking tokens (stderr output) using `2>/dev/null` to avoid bloat

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Low

README.md:51

Standard device file access

"Analyze this Claude Code skill repository comprehensively..." 2>/dev/null

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Low

SKILL.md:19

Standard device file access

4. When continuing a previous session, use `codex exec --skip-git-repo-check resume --last` via stdi

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Low

SKILL.md:20

Standard device file access

5. **IMPORTANT**: By default, append `2>/dev/null` to all `codex exec` commands to suppress thinking

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Low

SKILL.md:27

Standard device file access

| Read-only review or analysis | `read-only` | `--sandbox read-only 2>/dev/null` |

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Low

SKILL.md:28

Standard device file access

| Apply local edits | `workspace-write` | `--sandbox workspace-write --full-auto 2>/dev/null` |

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Low

SKILL.md:29

Standard device file access

| Permit network or broad access | `danger-full-access` | `--sandbox danger-full-access --full-auto

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Low

SKILL.md:30

Standard device file access

| Resume recent session | Inherited from original | `echo "prompt" \| codex exec --skip-git-repo-che

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Low

SKILL.md:31

Standard device file access

| Run from another directory | Match task needs | `-C <DIR>` plus other flags `2>/dev/null` |

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Low

SKILL.md:54

Standard device file access

- When resuming, pipe the new prompt via stdin: `echo "new prompt" | codex exec resume --last 2>/dev

The skill instructs agents to redirect stderr to /dev/null for Codex commands. That can hide warnings, diagnostics, or safety-relevant output during automated command execution.

Detected Patterns

System reconnaissanceNetwork reconnaissance

Audited by: codex View Audit History →

Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report

Copy report link

https://skillstore.io/skills/softaworks-codex/audits/6?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/softaworks-codex/security.svg)](https://skillstore.io/skills/softaworks-codex?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/softaworks-codex?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/softaworks-codex/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/softaworks-codex.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

Academic citations (APA · BibTeX · CFF)

APA citation

softaworks. (2026). codex security audit report (audit version 6) [Author version unspecified]. Skillstore. https://skillstore.io/skills/softaworks-codex/audits/6

BibTeX citation

@techreport{softaworks-softaworks-codex-2026,
  author = {softaworks},
  title = {codex security audit report (audit version 6)},
  institution = {Skillstore},
  year = {2026},
  number = {6},
  url = {https://skillstore.io/skills/softaworks-codex/audits/6},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "codex security audit report (audit version 6)"
version: "unspecified"
type: report
authors:
  - name: "softaworks"
date-released: "2026-07-09"
url: "https://skillstore.io/skills/softaworks-codex/audits/6"
identifiers:
  - type: other
    value: "skillstore:softaworks-codex:audit:6"
    description: "Skillstore immutable audit report identifier"

Skillstore Score

Why this score Evidence Confidence: Medium

Architecture

Maintainability

Content

Community

Spec Compliance

What You Can Build

Repository Review

Ask Codex to inspect a repository and return prioritized engineering improvements.

Guided Refactoring

Use Codex to plan or apply refactors while Claude coordinates review and follow-up.

Resume Agent Work

Continue a previous Codex session without rebuilding the full task context.

Try These Prompts

Review This Repository

Use Codex to review this repository in read-only mode. Summarize the top issues and suggest practical next steps.

Plan A Refactor

Use Codex with high reasoning effort to plan a refactor for this module. Do not edit files yet.

Apply Focused Changes

Use Codex to implement the approved change in workspace-write mode. Show the changed files and any tests run.

Resume And Verify

Resume the last Codex session. Ask it to verify the previous changes and identify any remaining risks.

Best Practices

Start in read-only mode for analysis before allowing edits.
Ask for explicit approval before using full-auto or danger-full-access modes.
Review Codex output and local diffs before accepting changes.

Avoid

Do not run broad-access commands for simple review tasks.
Do not hide stderr when debugging failures or safety warnings.
Do not resume a session with sensitive prompts unless the user approves.

Frequently Asked Questions

What does this skill do?

It lets Claude Code run Codex CLI tasks and resume Codex sessions for coding workflows.

Does it require Codex CLI?

Yes. Codex CLI must be installed, available on PATH, and configured with valid credentials.

Can it edit files?

Yes, when workspace-write or broader sandbox modes are selected for an approved task.

Can it access the network?

It can when danger-full-access is selected. Users should approve that mode before use.

Why is stderr hidden by default?

The skill hides stderr to reduce context size, but debugging and warnings may need stderr output.

Is this safe for private repositories?

Use caution. Codex may process repository content, prompts, and command output through configured Codex services.

Developer Details

Author

softaworks

License

MIT

Skillstore revision

Version notice

The author did not declare a version.

Repository

https://github.com/softaworks/agent-toolkit/tree/main/skills/codex/

Ref

3e4b6c31a74a3bd1a291c98cf585d720cb9fbc88

Maintenance freshness

7/18/2026

Usage

11 downloads · 185 views

File structure

📄 README.md

📄 SKILL.md

codex

Install with my Agent

Agent-readable resources

Test it

Security Audit

Confirmed security concerns (5)

Risk Factors

Detected Patterns

Copy report link

Markdown badge

HTML badge

Embed card

APA citation

BibTeX citation

CITATION.cff

Skillstore Score

What You Can Build

Repository Review

Guided Refactoring

Resume Agent Work

Try These Prompts

Best Practices

Avoid

Frequently Asked Questions

Developer Details