agentleFS
SuperJJ007SuperJJ007/papermachine/snapshots/AGENTS.md

papermachine / snapshots

AGENTS.md59 starsChanged 50 days agoMITView on GitHub

This tree contains only tests whose committed session JSONL is replay input and expected persisted output. Keep non-session ARIA, geometry, generator, CLI, and unit expected output with its owning app, script, or package; use test:expected, test:web, or test for its owning tier. Every process under test starts through the dsh CLI with a shipped profile and optional scenario patches. Test clients may drive a public protocol or browser interface; do not add another application entrypoint, hidden CLI mode, or executable…

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

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every process under test starts through the dsh CLI with a shipped profile and optional scenario patches.

    line 5

≈ “Common” is an estimate: at least 1 in 5 of the 36,368 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.md — Recorded-session snapshots

This tree contains only tests whose committed session JSONL is replay input and expected persisted output. Keep non-session ARIA, geometry, generator, CLI, and unit expected output with its owning app, script, or package; use test:expected, test:web, or test for its owning tier.

Every process under test starts through the dsh CLI with a shipped profile and optional scenario patches. Test clients may drive a public protocol or browser interface; do not add another application entrypoint, hidden CLI mode, or executable scenario driver.

Each scenario owns or explicitly references one primary Session role plus contiguous child roles. Canonical parent filenames are session[.vN].jsonl; children are session.<ordinal>[.vN].jsonl; v0 omits .v0, positive versions use lowercase .vN, and every filename agrees with its header. A directory may retain several generations of one role, but replay, record, and refresh select the numerically highest. Record and refresh write version-named outputs without renaming or deleting a committed generation; reviewed source-tree curation is separate. The owner alone records or refreshes the selected role. For an ordinary one-shot case, derive the user task and replay script from that selected JSONL; do not duplicate them in an input.json. Shared references are read-only, acyclic, point to the owner's selected parent generation, and are used only when another interface intentionally renders the same recorded behavior.

An owner that must keep a historical generation declares its exact sessionFormat.version and closed migration coverage names in snapshot.yml; record and refresh never rewrite that Session fixture. Owners without this declaration track current-writer output.

Committed sessions are normalization fixed points. Replace volatile identities with typed relationship-preserving tokens, replace request system prompts and tool schemas with tokens, and keep exactly one readable sidecar owner per header class. Never redact arbitrary user or tool text merely because it resembles an identifier.

An adapter-local symlink may expose a cross-profile prompt or schema sidecar only when snapshot.yml names that source; the corpus gate resolves the link and checks the declared target. The required snapshot lane runs these aliases on macOS and Linux.

Workspace seeds stay scenario-local. A scenario that mutates the workspace sets workspace.final: true and commits the complete result under workspace.expected/; use only the ignored .empty marker for an empty result. Record and refresh do not rewrite this independent oracle. Model prose and tool-result text do not prove the external effect.

pnpm run test:snapshot replays without writes. Recording and refresh use the explicit snapshot scripts, and every resulting JSONL, prompt, schema, protocol, UI, and workspace diff is reviewed before commit.

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.

Reviews can't be read right now.

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.