Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add coredo-eu/codex-claude-orchestrator --skill claude-pty-agentsgit clone --depth 1 https://github.com/coredo-eu/codex-claude-orchestratorWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/coredo-eu/codex-claude-orchestrator/claude-pty-agents)<a href="https://agentmods.dev/skills/coredo-eu/codex-claude-orchestrator/claude-pty-agents"><img src="https://agentmods.dev/badge/skills/coredo-eu/codex-claude-orchestrator/claude-pty-agents/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/coredo-eu/codex-claude-orchestrator/claude-pty-agents"><img src="https://agentmods.dev/badge/skills/coredo-eu/codex-claude-orchestrator/claude-pty-agents.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00105 | $0.04392 |
| Opus 5 | $0.00053 | $0.02196 |
| Sonnet 5 | $0.00021 | $0.00878 |
| Haiku 4.5 | $0.00011 | $0.00439 |
Grade A, and why
claude-pty-agents scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 399 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Claude PTY agents
Use persistent Claude Code processes owned by this Codex thread. Keep Codex as the owner of intent, material architecture or product tradeoffs, authority, conflicts, independent verification, and the final verdict. Treat this skill as transport and custody policy, never as additional authority.
Contract and boundaries
Give an edit-capable worker one compact contract:
Outcome: state that must become true.Done when: observable acceptance criteria.Boundaries: exact root, side effects, prohibited actions, and ownership.Authoritative context: applicable source-of-truth material and unknowns.Non-goals: adjacent work not to absorb.Known evidence: concise observed facts and material uncertainty.Required handoff: material evidence, risk, uncertainty, missing authority, deliberate non-actions, and custody needed for Codex to decide.
The worker is permanently local-only. It cannot commit, push, publish, deploy, control services, send external messages, administer the host, operate on credentials, modify Claude/Codex configuration, or perform destructive remediation. These restrictions bind the worker, not the owning Codex session. After custody returns, Codex may perform shared or external actions already authorized by the active goal and its unambiguous scope. Codex asks the user only for a material scope expansion or when the target or intended end state cannot be determined safely. Maintain one edit owner per canonical worktree.
The generated settings, deny rules, prompt, and hook are cooperative controls, not an OS sandbox. Bash or a malicious repository instruction can bypass path rules. Use Codex sandbox/approval policy, source review, and least privilege as the actual containment boundary.
Start or resume
Resolve this skill's directory, then check prerequisites and status:
<skill-dir>/scripts/toggle-agents.zsh status
command -v claude jq zsh git
Read status literally: busy is live active plus live reserved capacity;
active and reserved are durable open records; orphaned is a proven-dead
active writer; stale_reserved is a proven-dead reservation that ordinary
admission may safely tombstone; and blocked_roots includes every open record.
What ships with it
23 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- agents/openai.yaml 233 B
- assets/native-agents/mech_executor.toml.in 659 B
- assets/native-agents/reviewer.toml.in 472 B
- assets/native-agents/scout.toml.in 897 B
- assets/native-agents/security_reviewer.toml.in 494 B
- assets/native-agents/source_explorer.toml.in 496 B
- assets/native-agents/test_runner.toml.in 526 B
- assets/worker-agents.json 5.0 KB
- assets/worker-system-prompt.txt 2.0 KB
- references/codex-policy-snippet.md 5.4 KB
- scripts/assign-worker.zsh 12 KB runs code
- scripts/launch-worker.zsh 39 KB runs code
- scripts/reconcile-orphan.zsh 5.7 KB runs code
- scripts/retire-native-fallback.zsh 3.7 KB runs code
- scripts/rotate-worker.zsh 4.0 KB runs code
- scripts/run-native-agent.zsh 4.4 KB runs code
- scripts/runtime-lib.zsh 26 KB runs code
- scripts/setup-native-agents.zsh 6.7 KB runs code
- scripts/toggle-agents.zsh 7.0 KB runs code
- scripts/worker-agent-router.zsh 1.1 KB runs code
- scripts/worker-compaction-counter.zsh 724 B runs code
- scripts/worker-stage-guard.zsh 15 KB runs code
- scripts/worker-subagent-contract.zsh 809 B runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 399 lines · 105 tokens per session scan A efd72abd923c
claude-pty-agents is a skill published in the GitHub repository coredo-eu/codex-claude-orchestrator (5 stars, last pushed 5d ago), licensed MIT. It adds 105 tokens to every session and 4,392 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
coding-agents-farm
To orchestrate parallel coding-agent farms (Claude, Codex, Copilot, Gemini, etc.) on isolated git worktrees.
team-mode
A workflow for splitting substantial work into smaller independent tasks handled by specialized subagents. It coordinates exploration, implementation, and review while keeping final decisions in the main session.
allinluna
Compile and execute a goal, existing plan, active run, or Research Routes packet through the vNext Coordinator, Task Lanes, WorkUnits, and Workflow Packs.
squads-cli
Squads CLI — operate your AI workforce through the loop; intent in, reviewed work out. TRIGGER when running squads, dispatching agents, triaging the inbox, reading/writing memory, recording feedback, or checking status/usage.
fable-thinking
Reasoning protocol distilled from Claude Fable 5.1. Makes any model reason like Fable — evidence-grounded claims, multi-hypothesis diagnosis, concrete simulation, adversarial self-review, calibrated outcome-first delivery. Its never-skipped Floor check catches simple-looking trick questions models answer confidently…
squads-learn
Capture learnings after completing work. Use when finishing a task, fixing a bug, discovering a pattern, or learning something worth remembering for future sessions. Helps build institutional memory.