AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
Stop Babysitting LLMs, Start Shipping. Encode your existing SDLC into a gated multi-agent flow so LLMs do the noisy work.
Developer toolsHTML · RustGitHub @EffortlessMetricsOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
No updates detected yet. GitHub releases will appear here.
Owners of public repositories under a personal GitHub account can sign in to verify ownership and manage the introduction themselves. No administrator approval is needed.
Sign in, generate a verification file, and commit it to this repository’s default branch within 30 minutes. Files on other branches or forks do not count. Then check the file below to manage the project without administrator approval. Keep the file to retain your access.
If needed, you will go to sign-in. After signing in, return here and press your chosen verification button again.
Run /nomorevibe verify in your project folder to verify ownership and manage the introduction yourself.
Verify in one minuteAsk about other verification options
DemoSwarm template for Claude Code: governed SDLC flows producing review-ready PRs with evidence and receipts.
Taken from the repository README
Loading the full README.
Read the full README ↗A reusable SDLC pack for Claude Code: investigate a problem, build and challenge a coherent candidate, carry executed evidence into review, qualify deliberately, and make a separate integration decision.
The goal is useful software with less rework and operator effort. Faster code generation alone does not establish quality; neither do green panels or receipts without meaningful tests and current source.
/customize-pack
/flow-1-signal "Add a health check endpoint"
Continue through the relevant flows, or enter an existing issue/PR at its first missing or stale judgment. Read CLAUDE.md for the operating contract and late qualification for migration status. The installed consumer's workflow and protection remain authoritative; this instruction update does not silently change their triggers.
| Entry | Outcome |
|---|---|
| Signal | Investigated problem, intended behavior and falsifiers |
| Plan | Challenged design and bounded acceptance/rollback claim |
| Build | One coherent candidate, focused proof and quality improvements |
| Review | Substantive findings, verified repairs and current dispositions |
| Gate | Explicit independent qualification and its actual result |
| Deploy | Fresh final judgment, protected merge and landed reconciliation |
| Wisdom | Evidence-backed product and operating-method improvement |
These are recurring questions and useful entrypoints, not a mandatory sequence of fresh agents. Keep useful writer and build context warm. Different readers, tools and oracles can challenge the same candidate concurrently.
A Draft exposes source and evidence. Draft/Ready is not a qualification request. Source review can complete while independent CI is still NotRun. An explicit request binds the candidate, reviewed evidence and selected obligations; an old label cannot authorize a successor. Qualification does not merge or release. The historical Deploy command name remains for compatibility, but merge, tag, public release and production deployment are different authorized operations.
The PR description should help a reviewer locate the claim, actual source, executed proof, material findings, limitations and next action. It is not a substitute for inspecting load-bearing code, real consumers and plausible wrong behavior. A clean first review does not suppress later substantive investigation.
Git owns versioned source and decisions. GitHub owns current issue/PR/review/check
and merge facts. .runs/<run-id>/ retains useful local evidence and helper
compatibility; it is not a second live-state authority. Summaries do not multiply
the number of independent observations.
Flow Studio is the related flow visualization/harness project. Its implementation and distribution roles remain separate; updating this pack does not activate or release that system.
| Location | Content |
|---|---|
.claude/commands/ | Flow entrypoints |
.claude/agents/ | Optional focused workers and reviewers |
.claude/skills/ | Native helpers and task procedures |
.claude/rules/ | Loaded operating guidance |
docs/ | References, teaching material and historical rationale |
bash .claude/scripts/pack-check.sh --no-color
Use the actual repository's validation requirements. Instruction/source checks, harness behavior, downstream adoption and installed-product proof are different acceptances. Issue #22 retains the unfinished historical-guide, helper and multi-harness migration; no complete downstream CI cutover is claimed here.
Created by Steven Zimmerman (@EffortlessSteven). Licensed under Apache-2.0.