mancitrus/system-prompts-and-models-of-ai-tools
1
0
Fork 0
mirror of https://github.com/x1xhlol/system-prompts-and-models-of-ai-tools.git synced 2026-06-17 23:09:35 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
662 Commits 1 Branch 0 Tags 81 MiB
2f376ed97a
Commit Graph

4 Commits

Author SHA1 Message Date
Claude
ba5cd75466
docs(governance): replace CLAUDE.md with discovery-phase constitution v1.0.0 
Customer-triggered by: N/A (governance infrastructure)
Allowed-type: 3.7 (Documentation of Existing Behavior)
Truth-registry-updated: no
Claims-registry-updated: no

Replaces the generic project-context CLAUDE.md with a 16-section
discovery-phase operating constitution that constrains all coding agents
during Weeks 4-12:

- §2: Phase Gate definition (6 criteria, all must be Green)
- §3: 8 narrow allowed work types (bug fixes, security, V-tasks, scaffolding)
- §4: 12 explicit prohibited categories with refusal templates
- §5: 4 response templates for common founder requests
- §6: Pre-commit checklist with structured commit message format
- §7: 10 Arabic-first invariants
- §8: 7 evidence-first invariants
- §9-10: Truth Registry + Claims Registry integration rules
- §11: Override protocol when founder contradicts pre-committed decisions
- §12: External consulting document filter
- §13: Execution log format with N/A red-flag detection
- §14: 8 escalation triggers
- §15: Meta change protocol (formal decision + PR + version bump)
- §16: Quick response index lookup table

Also updates execution_log.md with Phase 2 Waves entries per §13 format.

Gates: architecture_brief 40/40, release_readiness 102/102, truth audit 19/19.

https://claude.ai/code/session_01W1rJthWDkasijTdXCfxVHs
 2026-04-17 11:44:52 +00:00
Claude
40ab7b86c2
feat(dealix): Phase 1 completion + Phase 2 foundation scaffolded 
PHASE 1 COMPLETION:

TASK-005 — Live gitleaks scan:
  Scanned 146 commits with gitleaks v8.20.1
  Result: 1 finding — FALSE POSITIVE (model name llama-3.1-70b-versatile
    in test fixture, not an API key)
  Added to .gitleaksignore
  rotation_log.md updated with scan results
  VERDICT: No real secrets in git history — repo clean for extraction

TASK-006 — Legal templates (bilingual):
  docs/legal/templates/IP_ASSIGNMENT_AGREEMENT.md — bilingual IP assignment
  docs/legal/templates/PRIVACY_POLICY_EN.md — PDPL/GDPR-aware template
  docs/legal/templates/PRIVACY_POLICY_AR.md — Arabic privacy policy
  docs/legal/templates/TERMS_OF_SERVICE_EN.md — SaaS ToS with MENA pricing
  docs/legal/templates/DPA_EN.md — Data Processing Agreement with annexes
  All marked as "DRAFT — must be reviewed by Saudi counsel before use"

TASK-006 — Trademark Filing Kit:
  docs/legal/templates/TRADEMARK_FILING_KIT.md
  Covers: DEALIX (Latin) + ديلكس (Arabic) + logo
  Classes 9, 42, 35 across KSA, UAE, Egypt, Jordan, Kuwait
  Application text ready to paste into SAIP + equivalents
  Agent recommendations (AGIP, Saba, Bird & Bird, Al Tamimi)
  Budget: ~90-120K SAR for full MENA coverage

Founder Decision Package:
  FOUNDER_DECISION_PACKAGE.md — single file with 4 decisions:
    1. GitHub org name (recommend: dealix-io)
    2. Entity structure (MISA vs DIFC vs ADGM)
    3. Saudi counsel engagement (15-30K SAR)
    4. Trademark filing (30-50K SAR initial)
  Total founder time to unblock: ~1 week + ~50K SAR

PHASE 2 FOUNDATION:

DEALIX_PHASE2_BLUEPRINT.md — 18-month category leadership plan:
  10 parallel streams (Frontend, Product, AI, Enterprise, Integrations,
    Scale, Commercial, Customer Platform, Trust, Category POV)
  Executable NOW vs Requires External Services vs Wait-for-PMF
  Phase 2 completion criteria (NPS >=50, NRR >=120%, etc.)

TASK-F201 — Design System foundation (scaffolded):
  packages/design-system/tokens/primitive.json — W3C Design Tokens format:
    Brand palette (50-900), neutral (50-950), critical/warning/success/info
    Space, radius, motion (duration + easing) tokens
    Typography with Arabic fontFamily + arabic-adjustment (1.15) for size
    Arabic line-height (1.8) for diacritics
  packages/design-system/tokens/semantic.json — light + dark themes:
    surface, fg, border, interactive, status semantic layers
  packages/design-system/README.md — principles + integration guide

TASK-CAT1340 (prep) — @dealix/arabic-ui package (scaffolded):
  packages/arabic-ui/src/normalize.ts:
    Diacritic-insensitive search (fatha/kasra/damma stripped)
    Hamza variants normalized (أ/إ/آ → ا)
    Waw-hamza, ya-hamza, taa-marbuta, alef-maksura handled
    arabicMatch() + arabicCompare() helpers
  packages/arabic-ui/src/numerals.ts:
    Western/Arabic-Indic/Eastern Arabic-Indic conversion
    formatCurrency() for SAR/AED/EGP/USD/JOD/KWD
    formatNumber() with locale awareness
  packages/arabic-ui/src/direction.ts:
    detectDirection() via Unicode bidi algorithm
    isolate() using U+2068/U+2069 for mixed-direction content
    isRTL() locale check
    hasArabic() presence check
  Future: release as OSS after 12 months of internal use

TASK-CAT1310 — Manifesto (bilingual draft):
  marketing/manifesto.md — 4 principles in Arabic + English:
    1. Arabic first, not Arabic translated
    2. Decisions backed by evidence, not opinion
    3. AI recommends, systems commit, humans approve
    4. Saudi compliance built-in, not bolted on
  Publication target: dealix.io/manifesto + dealix.io/بيان

TASK-CAT1320 — Dealix Labs (scaffolded):
  docs/labs/README.md — research program structure:
    Annual State of Arabic Enterprise AI report
    Quarterly Arabic LLM Benchmarks
    OTel semantic conventions proposal
    Open source: @dealix/arabic-ui + @dealix/design-system

TRUTH.yaml updated:
  Added Phase 2 capabilities section (all as 'partial' or 'roadmap')
  Added ISO 27001/17/18 and bug bounty to security_claims (all false)

All gates GREEN:
  Architecture Brief: 40/40
  Release Readiness Matrix: 71/71 (up from 53/53)
  Release Readiness Gate (blueprint): PASS
  Truth Registry Validator: VALID

https://claude.ai/code/session_01W1rJthWDkasijTdXCfxVHs
 2026-04-17 10:52:28 +00:00
Claude
fee51ffb06
feat(dealix): execute ALL automatable blueprint tasks 
TASK-001 (prep) — Repository Extraction Script:
  scripts/extract_dealix_repo.sh — automates git filter-repo extraction
  of Dealix-only paths to new GitHub org. Preserves commit history.
  Awaits founder decision on org name.

TASK-003 — Python Dependency Modernization:
  backend/pyproject.toml — full project spec with pinned versions:
    - fastapi, pydantic, sqlalchemy, asyncpg pinned
    - OpenTelemetry packages now included
    - pytest==8.3.4, pytest-asyncio==0.24.0 (stable)
    - Dev group with ruff, mypy, testcontainers
  Ready for uv sync to generate uv.lock.

TASK-004 — Node Dependency Hygiene:
  frontend/package.json — pinned packageManager=pnpm@9.12.0
    and engines.node >=20.10.0 <21.0.0

TASK-005 — Secrets Audit Infrastructure:
  .pre-commit-config.yaml — gitleaks + detect-private-key + detect-aws
    + ruff auto-fix + truth-registry-validator local hook
  docs/internal/rotation_log.md — rotation tracking template with
    scan commands (gitleaks, trufflehog3) and forbidden practices

TASK-006 — Legal Foundation Tracker:
  docs/internal/legal_status.md — tracks:
    - Company incorporation options (MISA vs DIFC vs ADGM)
    - IP assignment requirements
    - Privacy Policy / ToS / DPA review status
    - Trademark filing (KSA, UAE, Egypt, Jordan)
    - PDPL / ZATCA / NCA / SDAIA regulatory status
    - Professional indemnity + cyber + general insurance

TASK-010 (complete) — Truth Registry Tooling:
  scripts/validate_truth_registry.py — validates TRUTH.yaml structure,
    status values, and claims_registry.yaml alignment
  .github/workflows/truth-validation.yml — CI workflow on changes to
    truth registry or claims registry

TASK-101 — Release Readiness Gate (blueprint-spec):
  scripts/release_readiness_gate.py:
    - Required artifacts check (11 files)
    - TRUTH.yaml field validation
    - Forbidden claims scan in public docs
    - Architecture brief sub-gate
  Complements release_readiness_matrix.py (runtime checks).

Blueprint saved:
  DEALIX_EXECUTION_BLUEPRINT.md — authoritative execution doc

Updated:
  release_readiness_matrix.py — now 53/53 checks (was 41/41)
  docs/execution_log.md — full task tracking

All 3 gates GREEN:
  Architecture Brief: 40/40
  Release Readiness Matrix: 53/53
  Release Readiness Gate: PASS

Remaining P0 founder decisions (cannot be automated):
  - TASK-001: GitHub org name + run extraction
  - TASK-006: Entity incorporation + counsel engagement

https://claude.ai/code/session_01W1rJthWDkasijTdXCfxVHs
 2026-04-17 10:39:21 +00:00
Claude
020868a773
feat(dealix): TASK-999 State Audit + TASK-010 Truth Registry + Claims Registry 
TASK-999 — State Audit (docs/internal/STATE_AUDIT.md):
  Answered all 9 pre-execution questions with evidence:
  - Repo: still inside forked prompts repo (BLOCKER)
  - Tests: CI failing (dependency drift, not code)
  - RLS: migration exists, not applied to production
  - Idempotency: middleware exists, not in app stack
  - OTel: gateway spans only, packages not in requirements
  - Production: none, $0 infrastructure, $0 LLM, no customers

TASK-010 — Canonical Truth Registry (docs/registry/TRUTH.yaml):
  15 capabilities classified: 7 live, 4 partial, 4 roadmap.
  LLM policy, data residency, security claims all documented
  with honest status (soc2: false, rls: false, pdpl: in-progress).

TASK-010 — Claims Registry (commercial/claims_registry.yaml):
  8 approved claims (backed by runtime evidence)
  2 restricted claims (need qualifier)
  8 forbidden claims (never say: "enterprise-grade", "SOC 2 compliant",
    "better than Salesforce", "10x revenue", "full autonomy", etc.)

Execution log started at docs/execution_log.md.

https://claude.ai/code/session_01W1rJthWDkasijTdXCfxVHs
 2026-04-17 10:32:40 +00:00