ultra-builder-pro-cli
AGENTS.md2 starsChanged 12 days agoMITView on GitHub
UBP is one project method (ultra), a deterministic engine (ubp), and one advisory SessionStart hook for Claude Code and Codex. Intent is source code, the plan is compiled, progress is computed, and a change is a diff. The host model owns judgment; the engine owns progress and mechanical gates. No ACP runner, daemon or Python hook is shipped. - skills/ultra/SKILL.md: the host-neutral method. Its engine-path placeholder is replaced only in installed copies; there is no separate method source. - engine/FORMATS.md:…
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.
The file
We split this AGENTS.md into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 25
Never create or rewrite the user's global AGENTS.md or CLAUDE.md.
- Distinctive: line 29
Read the affected contract before editing.
- Distinctive: line 45
Commit, push, tag, publication and real host installation are separate effects requiring explicit authorization.
≈ “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.
Files
Ultra Builder Pro v6 Repository Guide
Product boundary
UBP is one project method (ultra), a deterministic engine (ubp), and one advisory
SessionStart hook for Claude Code and Codex. Intent is source code, the plan is compiled,
progress is computed, and a change is a diff. The host model owns judgment; the engine
owns progress and mechanical gates. No ACP runner, daemon or Python hook is shipped.
Sources of truth and layout
skills/ultra/SKILL.md: the host-neutral method. Its engine-path placeholder is replaced only in installed copies; there is no separate method source.engine/FORMATS.md: project file formats, evidence rules and exit codes.engine/*.cjs: the engine;bin/install.jsdispatches commands and installation.adapters/claude.js,adapters/codex.jsandadapters/_shared/: host packaging, managed ownership, provenance and doctor. Each plugin includes its own engine copy.adapters/_shared/runtime-assets.cjs: two hosts, one Skill and one hook.hooks/session-start.cjs: read-only status observation, always exiting zero..ultra/: project specification, checks, plan, rulings, ledgers and records..ultra/.runtime/is disposable. Git and these files carry project truth.tests/: engine and package contracts;adapters/_shared/tests/: shared install tests.
See docs/PLUGIN-ISOLATION-CONTRACT.md for ownership and activation boundaries.
Never create or rewrite the user's global AGENTS.md or CLAUDE.md.
Development and release checks
Read the affected contract before editing. Preserve protected acceptance files. Change the smallest authoritative source and every affected consumer. Run the narrow regression first, then the release checks; inspect the final diff and packed paths:
npm run verify:release
npm pack --dry-run --json
node bin/install.js --all --global --doctor --jsonThe release gate runs the remaining Node suites and the production dependency audit.
Validate changed Skills with the Skill Creator validator and changed Codex manifests with
Plugin Creator. Use temporary --config-dir directories for installation
tests; never use a real HOME for a test. Report unavailable checks and failed aggregate gates
without treating isolated retries as success.
Commit, push, tag, publication and real host installation are separate effects requiring explicit authorization. Use Conventional Commits; stage only authorized paths and omit AI co-author trailers. Do not change scope or acceptance to make a check pass.
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.

