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/noisy/noisy-coding/claude-mdgit clone --depth 1 https://github.com/noisy/noisy-codingWhat 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 | $0.00865 | $0.00865 |
| Opus 5 | $0.00432 | $0.00432 |
| Sonnet 5 | $0.00173 | $0.00173 |
| Haiku 4.5 | $0.00086 | $0.00086 |
Grade B, and why
noisy-coding CLAUDE.md scanned grade B with 2 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.
Sends data to an external URLmediumData exfiltration
A POST to an outside endpoint may be telemetry or may be exfiltration; either way the mod talks to somewhere, and you should know where.
curl -X POST http://127.0.0.1:7765/shutdown -d '{"delay_seconds":60}' Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -s http://127.0.0.1:8765/status | grep -o '"version": "[^"]*"' How it starts
The opening of the file, as written. The whole thing — 75 lines — stays where its author put it; the contents beside it link to each section on GitHub.
noisy-coding — agent notes
Voice coding for Claude Code: a daemon (production runs in Docker) plus Claude
Code hooks and an MCP server. Python backend in src/noisy_coding/, Vue
dashboard in dashboard/, hooks in hooks/.
Local development setup
Follow docs/local-development.md — do not improvise. In short: production
owns ports 8765–8767; the dev instance from this checkout runs on 7765
(scripts/dev_daemon.sh). Wiring a session to dev means BOTH:
- a second MCP server named
noisy-coding-dev(stdio, LOCAL scope — oneclaude mcp add --scope localper machine; NEVER commit it to a.mcp.json, the plugin auto-ships that file to end users), and - project-scoped hook overrides in this repo's
.claude/settings.jsonpointing at localhooks/*.pywithNOISY_CODING_LISTENER_PORT=7765. These DUPLICATE the global docker-exec hooks on purpose — global ones keep serving production, project ones serve dev.
Key docs
docs/hooks.md— the five hooks, why each exists, registration pathsdocs/ports.md— what each port is fordocs/local-development.md— dev instance next to production
Releasing
Run python3 scripts/bump_version.py X.Y.Z - it bumps all four version
files AND prints the release checklist. Follow it to the letter; the one
step agents keep forgetting: auto-generated release notes only list PRs
and miss direct commits - always write real notes (highlights for
humans) and publish with gh release edit vX.Y.Z --notes-file ... --draft=false,
then verify the GitHub release and the Docker image manifest.
Deploying to the local prod container is part of the release:
docker compose pull && docker compose up -d
curl -s http://127.0.0.1:8765/status | grep -o '"version": "[^"]*"'
The curl MUST report the version just released - up -d alone proves
nothing. docker-compose.yml runs the PUBLISHED image; never point it at
build: (a stale local build once shadowed a release as 2.13.4) - the
working tree runs via docker-compose.dev.yml under a distinct name.
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 · 75 lines · 865 tokens per session scan B 87f9fb795d53
noisy-coding CLAUDE.md is an instructions file published in the GitHub repository noisy/noisy-coding (6 stars, last pushed 3d ago), licensed MIT. It adds 865 tokens to every session, about $0.0043 per session on Opus 5. A static security scan graded it B with 2 findings (sends data to an external url, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
Diction AGENTS.md
Instructions for DictionLabs/Diction, covering diction gateway -- agent reference, routes, get /health, get /v1/models and post /v1/trial.
sayna CLAUDE.md
Instructions for SaynaAI/sayna, covering claude.md, project overview, development commands, feature flags and high-level architecture.
cadence-code AGENTS.md
Instructions for michael-L-i/cadence-code, covering agent guide, project overview, plugin layout, important paths and local commands.
speech-mcp AGENTS.md
Instructions for sandraschi/speech-mcp, covering agents.md — speech-mcp, project identity, architecture quick reference, ports and code rules.
cadence-code CLAUDE.md
Instructions for michael-L-i/cadence-code, covering claude code guide, project summary, claude-specific flow and working expectations.
speech-mcp GEMINI.md
Instructions for sandraschi/speech-mcp, covering speech-mcp: system context, 🛠 tech stack, 📡 registry & ports, 🏗 directory structure and 🧠 development standards.