igorkravcenkoigorkravcenko/snowshoe/AGENTS.md

snowshoe

AGENTS.md1 starsChanged 35 days agoApache-2.0View on GitHub

Entrypoint for agents working in this repo. Snowshoe (provisional) — Don't let your agents outrun your understanding. Keep your footing. A git-native personal comprehension map of a repository (Apache-2.0). Implementation status: HP1–4 CLI + tryable map UI/skill; routine ADRs A/B remain proposed. Install / PATH: snowshoe via @igorkravcenko/snowshoe; informal snow is not PATH. Skills/agents call snowshoe. Default signal is skill/human → CLI (work next gates init / routine refresh / advance). init does not install git hooks; opt-in hooks remain…

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 31,700 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 repository9 other files it gives its agents.
- .cursor/rules/agent-comms.mdcCursor rule
- .cursor/rules/project.mdcCursor rule
- brain-protocolSkill
- consistency-auditSkill
- grill-canonSkill
- grillingSkill
- grill-meSkill
- snowshoe-scratchSkill
- snowshoeSkill

The file

We split this AGENTS.md into the rules it gives an agent

6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Very specific:
Read docs/brain/START_HERE.md.
line 5
Distinctive:
Follow the read order there: CURRENT → ROADMAP → EXPERIMENTS.
line 6
Distinctive:
Treat CURRENT + accepted ADRs as truth.
line 8
Very specific:
No learning/quiz/verify, agent spawn, or fake features beyond the HP1–4 CLI + tryable map UI/skill unless CURRENT + an ADR say so.
line 18
Very specific:
Do not invent ARR, users, or competitors.
line 20
Distinctive:
Do not import trading / Nautilus / portfolio / bots from other repos.
line 21

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 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.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

AGENTS.md

AGENTS

Entrypoint for agents working in this repo.

Read docs/brain/START_HERE.md.
Follow the read order there: CURRENT → ROADMAP → EXPERIMENTS.
Load .cursor/skills/brain-protocol/SKILL.md before changing product claims or behavior.
Treat CURRENT + accepted ADRs as truth. Evidence, notes, and experiments are not.

Product (one line)

Snowshoe (provisional) — Don't let your agents outrun your understanding. Keep your footing. A git-native personal comprehension map of a repository (Apache-2.0). Implementation status: HP1–4 CLI + tryable map UI/skill; routine ADRs A/B remain proposed. Install / PATH: snowshoe via @igorkravcenko/snowshoe; informal snow is not PATH. Skills/agents call snowshoe. Default signal is skill/human → CLI (work next gates init / routine refresh / advance). init does not install git hooks; opt-in hooks remain intent (CLI not shipped). Stack + clarifications: ADR-2026-09-25-implementation-stack. Slices: ADR-2026-09-26-vertical-slice-cli, ADR-2026-09-27-map-ui-and-skill, ADR-2026-09-27-skill-drain-loop.

Hard rules

English for docs/brain/**, skills, rules, this file, commits, and PRs.
User-facing chat: English by default (see .cursor/rules/agent-comms.mdc). Maintainer Russian-chat preference is private overlay only.
No learning/quiz/verify, agent spawn, or fake features beyond the HP1–4 CLI + tryable map UI/skill unless CURRENT + an ADR say so.
CI: docs brain-docs lint (ADR-2026-09-25-brain-docs-ci) plus app CI (typecheck, bun test, Biome) (ADR-2026-09-27-app-ci-biome). No lefthook / husky / pre-commit yet. Product git-hook CLI is not shipped.
Do not invent ARR, users, or competitors. Unknown stays unknown.
Do not import trading / Nautilus / portfolio / bots from other repos.
Quality: Chief of Staff owns consistency audit on docs PRs before merge (.github/agents/auditor.agent.md / grill-canon). Agents draft packages carefully; do not race a shared local clone.

Skills

Skill	When
`brain-protocol`	Session start; any canon or behavior change
`grilling`	Stress-test a plan or decision (rounds + frontier)
`grill-me`	User says "grill me" / wants to be interviewed
`grill-canon`	Audit CURRENT/ADRs for staleness or contradiction
`snowshoe`	Product skill at `skills/snowshoe/` (Cursor: `.cursor/skills/snowshoe` symlink). Gate: PATH `snowshoe` (`install.md` only if missing), then `drain.md` (`work next`) or `learn.md`. Working repo without skill files: `snowshoe skill install --json --skills-path <harness-skills-dir>`
`snowshoe-scratch`	Satellite skill at `skills/snowshoe-scratch/` (Cursor: `.cursor/skills/snowshoe-scratch` symlink). Path without the map: `~/.snowshoe/learn/<sha256(utf8(realpath))>`; optional atlas is `map status --json --cwd` only (optional `help --json`). Do not drain, init, or write the ledger. Working repo: `snowshoe skill install` copies it with the product skill.
`.github/agents/auditor.agent.md`	Read-only consistency audit (CoS before merge)

Maintainer-only overlay lives in the sibling private repo snowshoe-maintainers. Link with ./scripts/link-private-overlay.sh (or set SNOWSHOE_MAINTAINERS):

.cursor/skills/gan-* — Orca GAN skills
docs/private — GTM / monetization research / competitive evidence / naming history
.cursor/rules/maintainer-chat.mdc — optional Russian chat preference

Symlinks are gitignored. Public CURRENT/ADRs must not depend on docs/private links. Product CLI/map does not depend on the overlay.

Download now (ZIP)

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

What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free