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/meltforce/vimmary/claude-mdgit clone --depth 1 https://github.com/meltforce/vimmaryWhat 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 | $0.06226 | $0.06226 |
| Opus 5 | $0.03113 | $0.03113 |
| Sonnet 5 | $0.01245 | $0.01245 |
| Haiku 4.5 | $0.00623 | $0.00623 |
Grade A, and why
vimmary CLAUDE.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 — 456 lines — stays where its author put it; the contents beside it link to each section on GitHub.
vimmary — video and podcast summary service
A Go service that turns bookmarked YouTube videos and transcribed podcast episodes into searchable LLM summaries. Video transcripts come from YouTube's InnerTube API, podcast transcripts from cast2md, summaries from Claude or Mistral, storage is Postgres + pgvector. Triggers are Karakeep webhooks, manual URL submission, cast2md feed subscriptions and a deep link from cast2md; results are served over a web UI, an MCP endpoint and three Atom feeds.
It is not a transcription service — the domain is classic YouTube videos (talks, tutorials) plus podcast episodes cast2md has already transcribed. Livestreams, Shorts and playlists are out of scope by decision, and vimmary never downloads audio for podcasts.
Gotchas
web/dist/ is generated and embedded into the binary. web.go carries
//go:embed all:web/dist, so go build ./cmd/vimmary fails outright when the
directory does not exist — this is what a fresh checkout looks like, because
.gitignore excludes it. Either build the frontend first (cd web && npm ci && npm run build) or create the stub the CI creates:
mkdir -p web/dist && touch web/dist/.gitkeep.
Two compose files, different audiences. docker-compose.yml is the
development stack and pulls :edge from a tailnet-internal registry, which is
unreachable from outside the tailnet. compose.example.yml is the one the
README points the public at; it pulls ghcr.io/meltforce/vimmary:latest.
Changing one does not change the other.
Nothing in the startup path fetches a secret over the network, and that is the
point. The init order in cmd/vimmary/main.go is config → tsnet → resolve the
database password from the environment → migrations → DB → services → HTTP
listener, and run() is that order — each step is a named function
(startTailscale, openDatabase, buildService, buildHTTPServer,
openListener), so the sequence is readable in one screen rather than inferred
from 254 lines. Every defer stays in run; moving one into the function that
created the thing it closes closes it immediately. The only remaining network
dependency is tsnet itself. Until
2026-08-07 vimmary read three secrets from setec over the tsnet node during
startup, and a node the tailnet had not yet placed got access denied — an
answer the setec store retries forever. That produced a 6h23min outage with the
container up and no listener; see INCIDENTS.md. Reintroducing any network call
before the listener reintroduces that failure class.
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 · 456 lines · 6,226 tokens per session scan A 55a2e4535a9c
vimmary CLAUDE.md is an instructions file published in the GitHub repository meltforce/vimmary (5 stars, last pushed 4d ago), licensed MIT. It adds 6,226 tokens to every session, about $0.0311 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
Pake AGENTS.md
AGENTS.md instructions for tw93/Pake, covering agents.md - pake project knowledge base, project identity, repository structure, development commands and code conventions.
Pake copilot-instructions.md
Copilot instructions for tw93/Pake, a project described as: 🤱🏻 Turn any webpage into a desktop app with one command.
Pake CLAUDE.md
Claude Code instructions for tw93/Pake, a project described as: 🤱🏻 Turn any webpage into a desktop app with one command.
SimpMusic CLAUDE.md
Claude Code instructions for maxrave-dev/SimpMusic, covering claude.md - simpmusic project guide for ai agents, 🌐 language rule, 📋 project overview, main purpose and basic information.
metube AGENTS.md
AGENTS.md instructions for alexta69/metube, covering agent guidelines, project scope — read this before planning a feature, readme.md size constraint, tech stack and build & test commands.
yutu AGENTS.md
Instructions for eat-pray-ai/yutu, covering yutu, quick reference, directory index, documentation and conventions.