Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsBlogPro
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges
  • Chrome Extension
  • Skill Manager

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Claude Delegate

ASecurity

Delegate a bounded task from Codex to the local Claude Code CLI. Use when the user explicitly asks Claude to investigate, implement, rescue, or continue work. Defaults to read-only; write delegation is confined to a generated git worktree.

8 stars
0 votes
0 copies
0 views
Added 9/24/2026
ai-agentsrustbashnodegit

Works with

claude codecli

Security Analysis

A100/100

Pro scans all 3 files and shows the line behind each finding

Scanned 9/30/2026

$npx -y skills add sanchitmonga22/cc-for-codex --skill claude-delegate --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Claude Delegate?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Claude Delegate
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/sanchitmonga22-claude-delegate/badge)](https://www.skillsdirectory.com/skills/sanchitmonga22-claude-delegate)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
Files
SKILL.md
---
name: claude-delegate
description: "Delegate a bounded task from Codex to the local Claude Code CLI. Use when the user explicitly asks Claude to investigate, implement, rescue, or continue work. Defaults to read-only; write delegation is confined to a generated git worktree."
---

# Claude Delegate

Full execution requests ultracode dynamic workflows with high model effort by
default. Only full-write delegation enables Workflow; `--no-ultracode` opts
out. Confirm an actual Workflow tool call before reporting orchestration as
successful. Workflow workers have full host access, not an OS sandbox.

Use the executable bridge launcher at the sibling skill path `../claude-code/scripts/cc-for-codex`, resolved to an absolute path. Invoke it directly, never through a bare `node` lookup.

## Modes

Ordinary Claude calls default to Claude Opus 5.5 (`claude-opus-5-5`) at
`high` effort. Explicit `--model claude-sonnet-5-5` selects Sonnet at
`high` effort. Full execution requests dynamic workflows by default; use `--no-ultracode` to opt out.
Do not route to Fable 5.1 unless the user specifically requests it.

- Foreground read-only: `delegate <prompt>`.
- Background read-only: after screening the prompt for secrets, disclose that background sessions have no supported max-budget guard and that the full prompt is temporarily visible to same-account local process inspection. After the user accepts both, add `--background --confirm-background unbounded-usage --confirm-background-data process-visible-prompt`.
- Foreground isolated write: the bridge defaults to `dangerous` zero-prompt permissions. Explain that a worktree is not a host sandbox and Claude file tools may reach host paths when permission checks are bypassed. After explicit authorization of both the edit and that risk, add `--write --confirm-write isolated-worktree --confirm-dangerous-permissions bypass-host-safety`.
- Foreground full-execution write: only when the user explicitly asks Claude to
  execute the task, add `--write --execution full --model claude-opus-5-5
  --effort high --confirm-write isolated-worktree
  --confirm-dangerous-permissions bypass-host-safety
  --confirm-execution full-host-access`. This adds Bash to the file tools and
  bypasses Claude prompts. It remains worktree-scoped by convention, not by OS
  sandbox; inspect all commands and the diff afterward.
- Foreground guarded write: when the user does not authorize bypass or requests the safer profile, add `--write --write-permissions guarded --confirm-write isolated-worktree`. This remains zero-prompt through `dontAsk` plus explicit `Edit,Write` preapproval, but protected Git/Claude/config paths are denied.
- Background isolated write: combine the chosen write-mode confirmation(s) with
  both background confirmations. File-only writes cannot run commands; full
  execution can run Bash after the additional `full-host-access` confirmation.
  Monitor the job with `$claude-sessions`, then inspect the worktree and have
  Codex validate afterward.

## One-writer rule

Never let Claude and Codex edit the same checkout concurrently. Write mode always generates a Claude worktree based on current local `HEAD`; it never reuses the Codex checkout, merges, pushes, cleans up, or deletes the worktree.

Before write delegation:

1. Confirm the current directory is a git repository.
2. For background/non-`-p` worktree delegation only, confirm the user has opened Claude Code interactively in this repository and accepted its workspace-trust prompt. Foreground `-p --worktree` skips that prompt. The bridge cannot accept trust for the user.
3. Inspect `.worktreeinclude` if present. It can copy ignored files and secrets. Pass `--confirm-worktree-include copy-ignored-files` only after the user accepts the exact contents.
4. Confirm the repository is a self-contained, full, non-sparse checkout and does not depend on Git alternates/`git clone --shared`, Git LFS, or another selected content filter. The bridge refuses alternate/partial/promisor repos, sparse checkout, and filtered tracked files rather than reading a sibling object store, fetching, running repository-controlled drivers, or checking out encoded/pointer content.
5. For dangerous mode, obtain explicit host-safety-bypass authorization before adding its exact confirmation token. Do not infer this from ordinary permission to edit files. The user's standing instruction in the current task may supply this authorization; do not ask repeatedly within the same delegated run.
6. Put the narrow task, acceptance criteria, excluded files, and required handoff in the prompt.
7. Stop editing the overlapping area until Claude finishes.

Afterward, use the wrapper-verified worktree path and branch printed by the command, inspect the diff yourself, run repository validation from an appropriate safe checkout, and keep implemented/tested/merged status separate.

Read [references/isolation.md](references/isolation.md) before any write or background delegation.

Attribution

sanchitmonga22sanchitmonga22
View sourceSee grades on GitHubMore from sanchitmonga22 →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Terse caveman voice: answer first, fluff gone, every technical fact kept. Use for /caveman, "caveman mode", "talk like caveman", "be brief", "less tokens". Stays on until "stop caveman" or "normal mode".

1100021 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

698621 votes

Writing Skills

Create and manage Claude Code skills in HASH repository following Anthropic best practices. Use when creating new skills, modifying skill-rules.json, understanding trigger patterns, working with hooks, debugging skill activation, or implementing progressive disclosure. Covers skill structure, YAML frontmatter, trigger types (keywords, intent patterns), UserPromptSubmit hook, and the 500-line rule. Includes validation and debugging with SKILL_DEBUG. Examples include rust-error-stack, cargo-dep...

3931 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3421 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Amp, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Grok Build, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

741 votes
View all in ai-agents →