spyglass
AGENTS.md0 starsChanged 25 days agoMITView on GitHub
Spyglass is a CLI that verifies code changes with real tools and keeps the proof after cleanup. An operator selects claims and judges behavior; the CLI records and validates the mechanical evidence. A successful command proves only the behavior it checked. TypeScript on Node 24, ESM, bundled by esbuild. Runtime dependencies are pg and zod; browser, Compose and secret tools stay external. This repository is public under MIT. Keep credentials, browser state, private application data, internal project names and personal…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 32,671 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository3 other files it gives its agents.
- create-verification-skillSkill
- maintain-verification-skillSkill
- spyglassSkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 35
Use npm ci for setup, npm run build for type-checking, bundles and generated schemas, npm test for E2E checks, and npm run format:check for …
- Distinctive: line 37
Read the testing guide before running or changing checks.
- Very specific: line 40
Use Git Flow: ordinary work branches from dev and returns through a pull request.
- Distinctive: line 47
Read the relevant documents before starting work.
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Spyglass
Spyglass is a CLI that verifies code changes with real tools and keeps the proof after cleanup. An operator selects claims and judges behavior; the CLI records and validates the mechanical evidence. A successful command proves only the behavior it checked.
Layout
TypeScript on Node 24, ESM, bundled by esbuild. Runtime dependencies are pg
and zod; browser, Compose and secret tools stay external.
src/: CLI, adapters, run state and offline report.schemas/: committed JSON Schemas generated fromsrc/schema.ts.tests/e2e/andfixtures/demo-app/: built-CLI checks and their real target.landing/: the static public landing page; no build step.skills/: source of truth for the three Spyglass skills; keep their licenses.bench/: scripted comparisons and approved public evidence.
Work boundaries
This repository is public under MIT. Keep credentials, browser state, private
application data, internal project names and personal machine paths out of Git.
Local, ignored .plans/ files can supply context, but shared instructions and
completion evidence must live in tracked documents. Read the current local plan
when present; never copy its private content into tracked files.
Preserve source identity, failed attempts and cleanup records. Generic modules must not contain integration paths or names from a target repository; the profile owns those details. Keep unrelated work unchanged and stage intended paths only.
Build and verify
Use npm ci for setup, npm run build for type-checking, bundles and generated
schemas, npm test for E2E checks, and npm run format:check for formatting.
Read the testing guide before running or changing checks. Implement a behavior
change only after its failure case and E2E check fail for the intended reason.
Use Git Flow: ordinary work branches from dev and returns through a pull
request. Read the workflow before branch, merge or release work. Publication
and release tags need explicit user approval. Follow docs/npm-publication.md
for npm staging, approval and registry installation checks.
Agent docs
Read the relevant documents before starting work. This file is the shared repository instruction source; task-specific rules stay in the documents below.
GLOSSARY.md: domain terms; read before naming concepts in code or prose.ARCHITECTURE.md: module ownership, run flow and invariants; read before changing structure, commands or state ownership.FRONTEND.md: offline report engineering; read before changing a browser-facing surface.DESIGN.md: implemented visual intent and surface exceptions; read before changing appearance or UI copy.agent_docs/testing.md: E2E setup, focused checks and failure-first changes; read before running or changing tests.agent_docs/git-workflow.md: branch roles, PR targets and protected merges; read before Git or GitHub delivery work.agent_docs/delivery.md: generated files, public contracts, scans and releases; read before changing output, pushing or releasing.agent_docs/ci.md: workflow gates, retained artifacts and failure diagnosis; read before changing CI or reviewing its evidence.docs/failure-cases.md: failures the CLI must detect; read before changing evidence, verification or cleanup behavior.docs/browser-benchmark.md: browser overhead budget and build choices; read before changingui, its proxy or the build.docs/verification-speed.md: speed targets and proof; read before changing tool driving, evidence recording or result validation.docs/roadmap.md: current priorities and completion limits; read before selecting milestone work.CONTRIBUTING.md: public contributor entry point; read before preparing a contribution.README.md,docs/quickstart.md,docs/usage.mdanddocs/commands.md: public command and operator contracts; read before changing the documented workflow.docs/configuration.md,docs/limitations.mdanddocs/support.md: public profile, limit and platform contracts; read before changing them.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

