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/talberthoule/backchannel/claude-mdgit clone --depth 1 https://github.com/talberthoule/backchannelWrote 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/instructions/talberthoule/backchannel/claude-md)<a href="https://agentmods.dev/instructions/talberthoule/backchannel/claude-md"><img src="https://agentmods.dev/badge/instructions/talberthoule/backchannel/claude-md.svg" alt="Measured on agentmods" height="20"></a>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 | $0.06410 | $0.06410 |
| Opus 5 | $0.03205 | $0.03205 |
| Sonnet 5 | $0.01282 | $0.01282 |
| Haiku 4.5 | $0.00641 | $0.00641 |
Grade B, and why
backchannel CLAUDE.md scanned grade B with 1 finding 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 today.
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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
Sentrux is installed at `C:/Users/thoule/.local/bin/sentrux.exe` and configured as a Codex MCP server in `C:/Users/thoule/.codex/config.toml`. How it starts
The opening of the file, as written. The whole thing — 337 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with this repository.
Coordination Record
The durable record for this repository is Linear: team Alpha, project Backchannel, issue IDs ALP-NNN. Findings, verdicts, decisions and their rationale, blockers, and declined alternatives go there; Herdr agent send carries only the issue ID and one line of what changed. Because the Linear MCP server writes with a single shared credential, every agent's issue and comment is authored by the human who owns the token, so record Requested by / Performed by / Date / Scope in the body. Branches are agent/alp-NNN-slug; design docs live in docs/superpowers/specs/ and plans in docs/superpowers/plans/. See the Routing Substance and Pointers section of the coordinating-herdr-agents skill.
Git worktrees live outside this checkout, at C:/work/backchannel/<lane>: git worktree add C:/work/backchannel/<lane> -b agent/alp-NNN-slug master. This repository sits in a OneDrive-synced folder, and sync repeatedly stripped the .git file out of worktrees created under .worktrees/, orphaning them from git worktree list while leaving the directories behind (ALP-139). Do not recreate .worktrees/ here. Note also that the Docker Compose stack binds the main checkout, so a worktree suits docs, review, and analysis rather than work needing integration testing.
Current Codebase Snapshot
Backchannel (formerly Call Helper) is a real-time meeting analysis app. A React frontend captures microphone audio (and optionally tab/system audio) as PCM16 16 kHz mono and streams it over WebSocket with a 1-byte track prefix. The FastAPI backend writes speaker-attributed transcript entries, records per-segment call audio to disk, runs provider-routed analysis agents (Gemini or OpenAI) over recent transcript text, and stores insights in PostgreSQL.
Backend tests live in backend/tests/ as stdlib unittest files; run them from backend/ with python -m unittest discover -s tests. Frontend behavior checks are npm run build (typecheck).
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.
- today Changed · -8 lines · -147 tokens per session 0e7ba186edff
- yesterday Changed · +13 lines · +758 tokens per session 8b06fc5c6cad
- 5d ago First seen · 332 lines · 5,799 tokens per session scan B b2479a7f8d96
backchannel CLAUDE.md is an instructions file published in the GitHub repository talberthoule/backchannel (5 stars, last pushed today), licensed MIT. It adds 6,410 tokens to every session, about $0.0321 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
CLIProxyAPI AGENTS.md
AGENTS.md instructions for router-for-me/CLIProxyAPI, covering agents.md, repository, commands, config and architecture.
vexa AGENTS.md
AGENTS.md instructions for Vexa-ai/vexa, covering agents.md — the agent front door (in-repo), researching vexa? start here, driving vexa from an agent session, your issue is your prd and claiming — and say hello on discord.
neo-chat AGENTS.md
AGENTS.md instructions for u14app/neo-chat, covering repository guidelines, project structure & module organization, build, test, and development commands, coding style & naming conventions and testing guidelines.
ima2-gen AGENTS.md
AGENTS.md instructions for lidge-jun/ima2-gen, covering ima2-gen — ai context, what this project does, tech stack, project structure and agent skills (packaged).
vexa CLAUDE.md
Claude Code instructions for Vexa-ai/vexa, a project described as: Open-source meeting transcription API for Google Meet, Microsoft Teams & Zoom. Auto-join bots, real-time WebSocket transcripts, MCP server for AI agents. Self-host or use hosted SaaS.