6.4 KiB
Prompt Policy and Change Control — Guardrails and Fix Patterns
A governance fix page for prompt stability, approval flows, and change control.
Use this page when uncontrolled prompt edits, hidden overrides, or missing approval gates destabilize your RAG or reasoning pipeline.
When to use this page
- Prompts are edited directly in production without sign-off.
- Fine-tuned prompts drift from documented baseline.
- Multi-agent pipelines apply different prompt rules with no central approval.
- No version history or rollback path exists for prompt changes.
- Waivers for unsafe prompt edits lack expiry or owner.
Acceptance targets
- Prompt policy coverage ≥ 0.95 across live agents, RAG steps, and evaluation sets.
- ΔS(question, retrieved) ≤ 0.45 after prompt edits (no semantic drift).
- λ_observe remains convergent across 3 paraphrases and 2 seeds.
- Each prompt edit has recorded owner, approval, and expiry date.
- Version rollback possible in under 60 seconds.
Typical breakpoints and WFGY fix
-
Live edits destabilize outputs
→ retrieval-traceability.md
Require citation schema and provenance anchors before prompt rollout. -
Hidden prompt injection bypasses control
→ prompt-injection.md
Enforce structural schema locks to prevent drift into unsafe states. -
Baseline prompts vanish after fine-tuning
→ policy_baseline.md
Require reference to immutable baseline for audit and rollback. -
Conflicts between agents (different policies in orchestration)
→ Multi-Agent Problems
Use role namespaces and explicit prompt slots to enforce alignment. -
No audit trail for prompt changes
→ audit_and_logging.md
Require immutable logging of who changed what, when, and why.
Minimal governance checklist
- Immutable baselines — Store canonical prompt versions in a controlled repo.
- Approval required — Every change must be signed off before rollout.
- Waivers expire — Unsafe or experimental prompts must auto-expire.
- Rollback path — 1-click restore of previous prompt version.
- Drift checks — Run ΔS/λ probes after every edit; block rollout if drift > 0.45.
- Audit logs — Immutable records joinable to governance lineage.
🔗 Quick-Start Downloads (60 sec)
| Tool | Link | 3-Step Setup |
|---|---|---|
| WFGY 1.0 PDF | Engine Paper | 1️⃣ Download · 2️⃣ Upload to your LLM · 3️⃣ Ask “Answer using WFGY + <your question>” |
| TXT OS (plain-text OS) | TXTOS.txt | 1️⃣ Download · 2️⃣ Paste into any LLM chat · 3️⃣ Type “hello world” — OS boots instantly |
🧭 Explore More
| Module | Description | Link |
|---|---|---|
| WFGY Core | WFGY 2.0 engine is live: full symbolic reasoning architecture and math stack | View → |
| Problem Map 1.0 | Initial 16-mode diagnostic and symbolic fix framework | View → |
| Problem Map 2.0 | RAG-focused failure tree, modular fixes, and pipelines | View → |
| Semantic Clinic Index | Expanded failure catalog: prompt injection, memory bugs, logic drift | View → |
| Semantic Blueprint | Layer-based symbolic reasoning & semantic modulations | View → |
| Benchmark vs GPT-5 | Stress test GPT-5 with full WFGY reasoning suite | View → |
| 🧙♂️ Starter Village 🏡 | New here? Lost in symbols? Click here and let the wizard guide you through | Start → |
👑 Early Stargazers: See the Hall of Fame —
Engineers, hackers, and open source builders who supported WFGY from day one.
⭐ WFGY Engine 2.0 is already unlocked. ⭐ Star the repo to help others discover it and unlock more on the Unlock Board.
