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/dougcalobrisi/erm/agents-mdgit clone --depth 1 https://github.com/dougcalobrisi/ermWrote 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/dougcalobrisi/erm/agents-md)<a href="https://agentmods.dev/instructions/dougcalobrisi/erm/agents-md"><img src="https://agentmods.dev/badge/instructions/dougcalobrisi/erm/agents-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 | $0.01174 | $0.01174 |
| Opus 5 | $0.00587 | $0.00587 |
| Sonnet 5 | $0.00235 | $0.00235 |
| Haiku 4.5 | $0.00117 | $0.00117 |
Grade A, and why
erm 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 5d 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 — 86 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — erm
erm is a local CLI that strips disfluencies (um, uh, er, erm, ah,
hmm, mhm, mm, uh-huh, and elongations) from English speech audio. This
file orients any AI coding agent (Codex, Copilot, OpenCode, Cursor, Gemini CLI,
pi.dev, Claude Code, …) working with the tool.
Full docs: https://doug.sh/docs/erm/ — source in
docs/. Always confirm exact flags witherm --help; never guess flag names or defaults.
Install
Requires Python 3.11+ and ffmpeg/ffprobe on PATH.
Prefer uvx (no persistent install; uv caches the env after first run). Where
uv isn't available, fall back to a venv:
uvx erm --help # tier 1: run straight from PyPI via uv
# fallback (no uv on PATH):
python3 -m venv .venv && source .venv/bin/activate && pip install erm && erm --help
In the iterate loop below, erm means the launcher you resolved: uvx erm …
under tier 1, or plain erm … after activating the venv.
ffmpeg: brew install ffmpeg / apt install ffmpeg / choco install ffmpeg.
Transcription runs on CPU by default; GPU needs CUDA runtime libs (--device auto
falls back to CPU).
Use — the iterate loop
erm input.wav --dry-run # inspect the cut-list JSON, render nothing
erm input.wav # write input-cleaned-<timestamp>.wav
erm validate input.wav output.wav # re-transcribe; assert no fillers survive
Render mode is the first decision:
--mode remove(default) — excises fillers; timeline shrinks.--mode silence— mutes in place; duration preserved (use for video A/V sync and multitrack stems).
Video inputs: a video file works like any other recording. By default erm
emits the cleaned audio only (.wav) — the common "pull the audio out of
this video" case. Add --video to render the picture too (container inferred
from the input; A/V stays in sync by construction). With --video, --mode silence stream-copies the picture losslessly, --video-splice {crossfade,cut}
picks the splice style, and --vcodec/--crf/--preset tune the re-encode.
See docs/video.md.
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.
- 5d ago First seen · 86 lines · 1,174 tokens per session scan A 982e4e1646bc
erm AGENTS.md is an instructions file published in the GitHub repository dougcalobrisi/erm (112 stars, last pushed 12d ago), licensed MIT. It adds 1,174 tokens to every session, about $0.0059 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 2026-08-30.
Other instructions, from other repositories
openscreen AGENTS.md
AGENTS.md instructions for getopenscreen/openscreen, covering agents.md, setup commands, development principles, project layout and code style.
watch-skill AGENTS.md
AGENTS.md instructions for oxbshw/watch-skill, covering working on watch skill, what this project is, layout, commands and rules that are not negotiable.
vibeframe AGENTS.md
Instructions for vericontext/vibeframe, covering agents.md, repository shape, commands, cli shape and agent workflow rules.
movi-player AGENTS.md
Instructions for MrUjjwalG/movi-player, covering agents.md — movi-player for ai coding assistants, 1. what this is, 2. package entry points, 3. architecture map and 4. the web component .
kinocut CLAUDE.md
Claude Code instructions for KyaniteLabs/kinocut, covering claude.md instructions for kyanitelabs/kinocut and empower orchestrator law.
kinocut copilot-instructions.md
Copilot instructions for KyaniteLabs/kinocut: Read the nearest AGENTS.md for repository rules. For Kinocut workflows, load skills/kinocut/SKILL.md; it is the canonical public skill. Keep this adapter limited to host-specific discovery and do not duplicate policy here.