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/craigjmidwinter/getvect/agents-mdgit clone --depth 1 https://github.com/craigjmidwinter/getvectWrote 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/craigjmidwinter/getvect/agents-md)<a href="https://agentmods.dev/instructions/craigjmidwinter/getvect/agents-md"><img src="https://agentmods.dev/badge/instructions/craigjmidwinter/getvect/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 | $0.00961 | $0.00961 |
| Opus 5 | $0.00481 | $0.00481 |
| Sonnet 5 | $0.00192 | $0.00192 |
| Haiku 4.5 | $0.00096 | $0.00096 |
Grade A, and why
getvect 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 4d 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 — 69 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
GetVect is a local desktop vectorizer: an Electron app (TypeScript; React
renderer, pure-Node engine in src/engine/) that turns PNG/JPEG/BMP into
SVG/EPS/DXF/PDF/PNG on the user's machine — no accounts, no server, offline by
default. The engine is deterministic and heavily instrumented; the measurement
harness, not opinion, decides whether a change ships.
Commands (green on a clean checkout)
npm install # ~10 s; postinstall de-quarantines Electron on macOS
npm run typecheck # tsc strict, both configs
npm run test:engine # the engine contracts, pure Node, ~20 s — the fast gate
npm test # engine contracts + full Playwright acceptance suite (~3 min)
npm run instruments # corpus metrics + threshold gates → artifacts/metrics.json
npm run assets # regenerate derived/demo assets (CI fails if stale)
There is no linter yet (tracked in katra/tasks/); tsc strict is the only
style gate. Match the surrounding code's comment density and idiom — comments
in this repo state constraints and reasoning, not narration.
Invariants a PR must not break
- The harness decides. Read
docs/HARNESS.mdbefore touching the engine. Corpus fixtures decide; in-house artwork (Frankie) reports but cannot carry a decision. A rate measured over a short boundary is not a defect. State the window of every measurement. - Determinism: same image + same settings = byte-identical SVG. Tested.
- Sharp corners stay sharp. The de-staircasing failure mode is rounding
what was meant to be angular; the mascot's ear tip is the canary and
tests/engine/sharp-corners.test.mjsenforces it. - DEAD GATE rule: a declared threshold whose metric is not produced fails the runner. Never leave a gate that cannot go red.
- The CLI exit codes and flags are a published interface, in the same class
as the on-disk katra format, not an internal enum. They are documented in
SKILL.mdanddocs/CLI.mdas a contract an outside caller can be written against. No such caller is known today: the reason to freeze the table is that nobody can build against one that moves, so stability is what makes an integration possible at all. Adding a code or flag is free; renumbering or removing one breaks a caller silently, since their branch just stops matching.skill-doc.test.mjspins both tables by name and fails even when the docs move in the same commit. Change them deliberately and with a note, never as a refactoring side effect. - Asset staleness gate: derived assets (screenshots, demo media, icons)
are regenerated by committed scripts only — never hand-edited; CI runs
regenerate-derived-assets.mjs --checkand fails on drift. - Exactly two network touchpoints, both in
src/main/: the once-per-launch release-feed check (opt-out viaGETVECT_NO_UPDATE_CHECK) and opt-in AI Enhance with a user-supplied key (encrypted via safeStorage, never returned to the renderer). Nothing else may touch the network; the renderer ships CSPdefault-src 'self'. - Provenance is enforced. Nothing enters
fixtures/without a stated redistributable licence (build fails). Never vendor, commit, or quote competitor output — this repo's history was rewritten once to purge exactly that. The devlog publishes only via the opt-in gate (scripts/devlog-gate.mjs); its forbidden-names list is load-bearing. - Tests are invisible. The acceptance suite drives real Electron windows
that are never shown (
z-window-guard.spec.tsenforces it). Never launch the app headed from an agent session.
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.
- 4d ago First seen · 69 lines · 961 tokens per session scan A 5534d3470308
getvect AGENTS.md is an instructions file published in the GitHub repository craigjmidwinter/getvect (90 stars, last pushed 7d ago), licensed MIT. It adds 961 tokens to every session, about $0.0048 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
opencove AGENTS.md
AGENTS.md instructions for DeadWaveWave/opencove, covering system prompt: opencove project ai developer agent, 1. core directives & golden rules, 2. decision framework (small vs large), a. small change (fast feedback / 小步快反馈) and b. large change (deep thinking / 慎重对齐).
wesight AGENTS.md
Instructions for freestylefly/wesight, covering claude.md, build and development commands, build production bundle (typescript + vite), lint with eslint and run memory extractor tests (node.js built-in test runner).
limboo CLAUDE.md
Claude Code instructions for limboo-ai/limboo, covering claude.md, 1. what is limboo?, 2. tech stack (current), 3. project structure and the three electron contexts (critical mental model).
anytype-ts AGENTS.md
AGENTS.md instructions for anyproto/anytype-ts, covering agents.md, project structure, state management with mobx, electron integration and inter-process communication (ipc).
open-codesign AGENTS.md
Instructions for OpenCoworkAI/open-codesign, covering agents.md - open codesign, what this project is, hard constraints, ai visibility for web work and current architecture direction.
CodePilot CLAUDE.md
Claude Code instructions for op7418/CodePilot, covering claude.md, 协作模式(作者 / claude code / codex), 开发规则, 语义验收与反假数据 and 自检命令.