🛡️ Simplicio-loop-quality — Verified quality for any repository

Not a product. Blocked until Loop exposes public quality hooks. Do not treat this repo as a sixth Simplicio module or a second scheduler. See wesleysimplicio/simplicio-runtime#4213 and docs/QUALITY_EXTENSION.md on Runtime.

Agent-driven, fail-closed testing and quality assurance executed by simplicio-loop.

TL;DR · How it works · 36 lanes · Run it · Star history · Contributing

🌍 Languages:
🇬🇧 English | 🇧🇷 Português | 🇪🇸 Español | 🇫🇷 Français | 🇩🇪 Deutsch | 🇮🇹 Italiano | 🇯🇵 日本語 | 🇰🇷 한국어 | 🇨🇳 简体中文 | 🇷🇺 Русский | 🇵🇱 Polski | 🇹🇷 Türkçe | 🇳🇱 Nederlands | 🇮🇳 हिन्दी | 🇸🇦 العربية

---

🚀 TL;DR

Simplicio-loop-quality is the quality extension for the Simplicio product line. It decides which quality checks apply, supplies declarative testing agents, validates evidence, and hands the final decision to the simplicio-loop Completion Oracle.

It is deliberately not a second orchestrator. The Loop remains the only owner of Hub slots, queues, scheduling, isolation, processes, retries, recovery, delivery, and terminal completion.

Project metadata

Description: Simplicio-loop-quality — agent-driven, fail-closed quality gates for any repository via simplicio-loop.

Topics: simplicio-loop-quality · simplicio-loop · testing · quality-assurance · quality-gates · test-automation · ai-agents · evidence-driven · fail-closed · software-verification

Current status: architectural bootstrap with a fail-closed gate. Production execution remains blocked until the required upstream Loop extension hooks are available.

🔭 How it works

flowchart LR
  REPO["Any repository"] --> EXT["Simplicio-loop-quality"]
  EXT --> POLICY["Policy + applicability matrix"]
  POLICY --> LOOP["simplicio-loop\n(single authority)"]
  LOOP --> HUB["Loop Hub slots\nand isolation"]
  HUB --> AGENTS["24 declarative\nquality agents"]
  AGENTS --> EVIDENCE["Bound evidence\nreceipts"]
  EVIDENCE --> GATE["Independent\nQuality Gate"]
  GATE --> ORACLE{"Completion\nOracle"}
  ORACLE -->|PASS| DONE["Verified delivery"]
  ORACLE -->|BLOCKED| FINDING["Structured finding\nfor recovery"]

Loading

Evidence is collected during the authoritative run, not written after the fact.

The extension contributes policy, agents, adapters, plans, findings, evidence and gates. It never creates a local scheduler, worker pool, process supervisor, worktree manager or alternative done state.

🤖 Quality-agent constellation

flowchart TD
  PLAN["Quality plan"] --> WAVE["Loop-managed wave"]
  WAVE --> TEST["Test author"]
  WAVE --> INTEG["Integration / system"]
  WAVE --> PERF["Performance / resilience"]
  WAVE --> SEC["Security / supply chain"]
  WAVE --> REAL["Real-runtime / packaging"]
  TEST --> AUDIT["Independent evidence auditor"]
  INTEG --> AUDIT
  PERF --> AUDIT
  SEC --> AUDIT
  REAL --> AUDIT
  AUDIT --> GATE["Fail-closed Quality Gate"]
  GATE --> ORACLE["Loop Completion Oracle"]

Loading

Agents may author evidence; an independent auditor decides whether that evidence is admissible.

🧪 The 36 quality lanes

The strict policy covers unit, component, integration, contract, system, end-to-end, regression, smoke, real-runtime, negative-path, property, fuzz, mutation, invariant, concurrency, fault-injection, repeatability, security, supply chain, performance, load/stress/soak, coverage, compatibility, installation, upgrade/downgrade, migration, operational readiness, privacy, documentation, observability, implementation completeness, code review, static quality and evidence audit.

Applicability is explicit: an inapplicable lane needs a technical reason and independent approval. Skipped, flaky, stale, missing or unverifiable evidence never becomes PASS. Missing metrics are null with a machine-readable reason, never zero or an estimate.

🗺️ Declarative flow (simplicio.flow/v1)

The product flow conforms to simplicio.flow/v1 and maps the complete processing lifecycle: inputs (CLI, extension provider, policy, repository workspace, receipts), stages across the 36 quality lanes, persistent contract/journal stores, and definitive gate verdicts/remediation outputs.

Artifacts in docs/flow/:

• docs/flow/simplicio-loop-quality.flow.json: Single declarative source of truth.
• docs/flow/simplicio-loop-quality.mmd: Derived Mermaid diagram.
• docs/flow/simplicio-loop-quality.svg / .png: Rendered visual charts.
• docs/flow/langflow/simplicio-loop-quality.langflow.json: Importable flow in Langflow 1.12.0.

Regenerate flow

# Regenerate flow.json, mmd, svg, png and langflow artifacts:
make flow
# or:
python scripts/generate_flow.py

# Validate drift between code and flow specification:
make check-drift
# or:
python scripts/generate_flow.py --check-drift

▶️ Run it

pip install -e .

simplicio-loop-quality doctor
simplicio-loop-quality agents
simplicio-loop-quality plan --repo /path/to/project --out quality-task.md
simplicio-loop-quality run --repo /path/to/project
simplicio-loop-quality gate \
  --receipt quality-evidence.json \
  --source-sha "$LOOP_SOURCE_SHA" \
  --artifact-root "$LOOP_ARTIFACT_ROOT"

run starts exactly one authoritative Loop process after doctor proves the required extension capabilities. The standalone gate command is diagnostic; authoritative completion requires the Loop-ledger provenance and Completion Oracle.

⭐ Star history

Track the project as the quality layer matures:

🧭 Scope boundary

Quality agents may create or improve tests, fixtures, test configuration, quality CI and evidence. A product defect becomes a structured finding for the Loop recovery/implementation stage; this extension does not implement unrelated product behavior.

See Architecture, Quality contract, and the upstream requirements.

🛠️ Development

python scripts/check.py
pytest --cov=simplicio_loop_quality --cov-branch --cov-fail-under=85

🤝 Contributing

Open an issue with a reproducible quality gap, the affected lane, evidence, and the expected acceptance criteria. Changes must preserve the single-authority Loop boundary and the independent evidence gate.

License

MIT