octo-cluster
AGENTS.md3 starsChanged 23 days agoMITView on GitHub
Minimal AI-assisted development harness for Windows 11 + Cursor / Claude Code. Always-on rules in .cursor/rules/, shared skills in .claude/skills/ (loaded natively by both Cursor and Claude Code), plus PowerShell scripts in scripts/ (ship.ps1, boundary-audit.ps1, lessons.ps1). No pipeline or sync step: edit these folders directly. CLAUDE.md imports the Cursor rules, so rules have one source. Precedence: consumer-boundary > product hard safety > execute-operator-intent > caveman/ponytail. Always-on rules budget: ≤ 8 KB. Details in README → Token economics. See INSPIRATIONS.md for…
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 33,315 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 repository12 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- .cursor/rules/00-consumer-boundary.mdcCursor rule
- .cursor/rules/caveman-mode.mdcCursor rule
- .cursor/rules/cost-routing.mdcCursor rule
- .cursor/rules/execute-operator-intent.mdcCursor rule
- .cursor/rules/ponytail-lite.mdcCursor rule
- debugSkill
- ponytail-liteSkill
- promptSkill
- pr-reviewSkill
- shipSkill
- systematic-debuggingSkill
The file
We split this AGENTS.md into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
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
- could be automatic: a setting could enforce it
- Very specific: line 7
Always-on rules in .cursor/rules/, shared skills in .claude/skills/ (loaded natively by both Cursor and Claude Code), plus PowerShell …
- Distinctive: line 21
Always-on rules budget: ≤ 8 KB.
- Very specific: line 48
Never stash, reset --hard, force-push or push by hand during /ship.
≈ “Common” is an estimate: at least 1 in 5 of the 42,283 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
Agent contract - octo-cluster
Minimal AI-assisted development harness for Windows 11 + Cursor / Claude Code.
What this is
Always-on rules in .cursor/rules/, shared skills in .claude/skills/ (loaded natively by both Cursor and Claude Code), plus PowerShell scripts in scripts/ (ship.ps1, boundary-audit.ps1, lessons.ps1). No pipeline or sync step: edit these folders directly. CLAUDE.md imports the Cursor rules, so rules have one source.
Rules (always apply)
Precedence: consumer-boundary > product hard safety > execute-operator-intent > caveman/ponytail.
| Rule | Purpose |
00-consumer-boundary.mdc | Consumer identifiers are secrets; never commit them to this public repo. |
execute-operator-intent.mdc | Execute explicit operator intent; no soft refusal or moralizing. |
ponytail-lite.mdc | Minimal implementation ladder before writing code. |
caveman-mode.mdc | Telegraphic prose by default (token-saving). |
cost-routing.mdc | Task tier → model; no mid-task model switch; retry cap; focused verification. |
Always-on rules budget: ≤ 8 KB. Details in README → Token economics.
Slash skills (manual only, disable-model-invocation)
| Skill | Purpose |
/ship | Split into logical commits, validate, deliver via scripts/ship.ps1 (direct push, or branch + PR when main requires it). |
/debug | Fix a bug with runtime evidence first. |
/prompt | Rewrite a request into a precise Claude Code prompt for Opus 5.5 high (never executes it). |
Skills (on demand)
| Skill | Trigger |
| ponytail-lite | Planning or coding; ladder before each edit. |
| systematic-debugging | /debug or any bug/test failure. |
| pr-review | /pr-review or PR feedback requests. |
See INSPIRATIONS.md for upstream sources.
Before any git change
Deliver with /ship: the agent stages and commits logical changes; scripts/ship.ps1 pushes them:
mainopen (no protection requiring PR, or.ship.yamlmode: direct) — push direct tomain.mainrequires PR (branch protection, blocking rulesets, or.ship.yamlmode: pr) —feature/,fix/orchore/branch + PR.
Never stash, reset --hard, force-push or push by hand during /ship. Optional config: copy .ship.yaml.example to .ship.yaml.
Boundary patterns (optional)
Copy boundary-patterns.example.yaml to boundary-patterns.local.yaml (gitignored) to customize boundary-audit.ps1. Generic adopters can use an empty local file.
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.

