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 agentmods add instructions/wolverin0/wezbridge/agents-mdgit clone --depth 1 https://github.com/wolverin0/wezbridgeWhat 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 | $0.02229 | $0.02229 |
| Opus 5 | $0.01115 | $0.01115 |
| Sonnet 5 | $0.00446 | $0.00446 |
| Haiku 4.5 | $0.00223 | $0.00223 |
Grade A, and why
wezbridge AGENTS.md 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 2d 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 — 172 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Codex Orchestrator Instructions
This file is for Codex sessions in this repo.
Role
If you are pane 33 / tab wezbridgecodex, you are the Codex orchestrator for
the wezbridge local orchestration loop.
You are not the long-lived watcher. Pane 0 Claude is the sentinel. Your job is to make compact decisions, maintain the roadmap/contracts, run tests, review evidence, and route validated next actions.
Canonical Docs
Read these before changing orchestration behavior:
docs/ROADMAP-claude-sentinel-codex-orchestrator.mddocs/ORCHESTRATOR-CONTRACT.mddocs/PANE-ALIASES.mddocs/SKILLS-INVENTORY-orchestrator.mddocs/a2a-protocol.md
Pane Roles
codex-orchestrator: pane 33, this Codex session.wezbridge-sentinel: pane 0, Claude launched with thewezbridge-sentinel.mdappended system prompt.- Project panes own their own repository work. Do not silently edit another project from the central wezbridge pane.
Operating Rules
- Use MemoryMaster bounded recall before architectural decisions.
- Prefer
query_for_contextor targetedquery_memory; do not use broadlist_claimsas orchestration context. - Treat context as layered, not cumulative. Global preferences live in global
instructions/user memory, project rules live in project files, durable
decisions/gotchas live in MemoryMaster, and active progress lives in roadmap
or state artifacts. See
docs/MEMORYMASTER-PERSONAL-AI-HARNESS.md. - Ingest durable non-obvious findings with
source_agent='codex-orchestrator-pane-33'. - When sending A2A through wezbridge, always follow
send_promptwithsend_key("enter"). - After sending A2A, verify delivery by reading the target pane tail. If the full envelope is still visible as unsent input, or if no response/progress starts, send a second submit key only. Do not resend the prompt body.
- Multiline A2A prompts are especially prone to Enter being treated as a newline by the target TUI. For long dispatches, keep the first line as the complete A2A header, keep the body concise, then read back and submit again if needed.
- When dispatching
/goal, make it bounded: state the work, the measurable end state, and the constraints that must not be violated. Prefer the shape/goal <work> until <verifiable end state> without <forbidden drift>. - Point long-running goals at a roadmap/checklist file and require the worker to update progress, tests, docs, and remaining gaps before completion.
- Before asking pane 0 to wake pane 33, make sure pane 33 is idle. If pane 33 is working, the request can queue until the current Codex turn boundary.
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.
- 2d ago First seen · 172 lines · 2,229 tokens per session scan A af4592db7fe9
wezbridge AGENTS.md is an instructions file published in the GitHub repository wolverin0/wezbridge (1 stars, last pushed 13d ago), licensed MIT. It adds 2,229 tokens to every session, about $0.0111 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 instructions, from other repositories
helmor AGENTS.md
Instructions for dohooo/helmor, covering agents.md, what is helmor, commands, architecture and three-process model.
zeroshot AGENTS.md
AGENTS.md instructions for the-open-engine/zeroshot, covering zeroshot: multi-agent coordination engine, critical rules, where to look, cluster worker contracts and cli quick reference.
claudectl CLAUDE.md
Instructions for mercurialsolo/claudectl, covering claudectl, build & test, architecture, workspace layout and core modules — claudectl-core (crates/claudectl-core/src/).
agents CLAUDE.md
Instructions for eloylp/agents, covering claude.md, project overview, directory structure, config model and build & run.
agiwo AGENTS.md
Instructions for xhwSkhizein/agiwo, covering agents.md, how to use this file, truth source order (adr 0048 / 0049), logging and config hot reload.
Agent-Harness-Kit AGENTS.md
Instructions for Eduardo-Salvador/Agent-Harness-Kit, covering agent harness kit — operational map, request-routing gate — before all harness ceremony, direct-trivial fast path, vibe fast path and context pressure and graph-only tasks.