Meta Platform Agent Prompt Api Catalog V2
Meta Platform Agent Prompt Api Catalog V2
Source: meta-platform-agent-prompt-api-catalog-v2.md (ingested 2026-03-28)
Summary of 269185 character source file.
Searching 13 files for "prompt"
Agent Prompt API System
Version: 2.0
Purpose: Systematic discovery and cataloging of all project resources via structured agent definitions
Total Agents: 120 specialized agents across 12 categories
Last Updated: 2025-01-27
C:\Users\mesha\Desktop\Desktop_Files\META-PLATFORM\Master Synthesis.md:
13 - governance: Tiered rules, compliance scoring, property-based targeting.
14 - gents: Function agents (code-gen, reviewer, tester, deployer, healer) + domain agents (Optimization/QAP, Physics, AI tools, Web, Data Science) coordinated via LangGraph.
15: - prompts: Centralized, provider-specific (claude, openai, shared), versioned with routing profiles.
16 - ools: Typed CLI (scaffold, sync, publish, manage) favoring progressive adoption and escape hatches.
17 - PROJECTS: Git submodules for local control and global governance.
..
20 - Tier 2 (~8%): Sonnet/4o for moderate complexity.
21 - Tier 3 (~2%): Opus/o1 for deep reasoning & critical changes.
22: - Fallback chain: Claude → GPT → Gemini → local; prompt adaptation per model and context-window awareness.
23 - Cost model: –,400/month vs ,000–,000 naive; batch APIs for non-urgent audits; regional routing only if residency mandates.
24 - Orchestration backbone: Dagster asset graph with dry-runs, staged rollouts, parallel DAG execution, retries/timeouts/fallbacks, state passing, conditional logic.
..
62 - Sync governance: meta-platform sync-governance --domain optimization --all-repos
63 - Compliance across submodules: meta-platform compliance-check --all-repos
64: - Prompt sync: meta-platform sync-prompts --all-repos
65
66 ## Quality, Redundancies, and Gaps
..
74 ## Cross-References
75 - Governance tiers ↔ Approval framework; dry-run diffs & staged rollouts feed decisions.
76: - Multi-LLM routing ↔ Agent metrics & cost dashboards; prompt versioning aligns to routing profiles.
77 - Template evolution ↔ Copier migration tests; Phase 2 v1→v2 migrations.
78 - Observability ↔ Grafana dashboards; circuit breakers use signals for self-regulation.
..
89 - Stand up minimal Dagster pipeline: governance scan → dry-run diff → approval decision → canary rollout → metrics emission.
90 - Define LangGraph skeleton for 2 function agents (code-gen, reviewer) and 2 domain agents (Python, Web); instrument metrics.
91: - Create prompts/claude/ routing profiles and caching policy; wire batch flows for audits.
92
API Overview
Master Synthesis Overview
- Governance: Tiered rules, compliance scoring, property-based targeting
- Agents: Function agents (code-gen, reviewer, tester, deployer, healer) + domain agents (Optimization/QAP, Physics, AI tools, Web, Data Science) coordinated via LangGraph
- Prompts: Centralized, provider-specific (claude, openai, shared), versioned with routing profiles
- Tools: Typed CLI (scaffold, sync, publish, manage) favoring progressive adoption and escape hatches
- Projects: Git submodules for local control and global governance
C:\Users\mesha\Downloads\CONSOLIDATION\RESEARCH_CORE.md:
353 ---
354
355: # 🤖 SUPERPROMPTS
356
357: ## MEZAN Code Review Prompt
358
359: Use this prompt to review and improve MEZAN algorithm selection code:
360
361 ... 399
400
401: ## Turing Challenge Validation Prompt
402
Base Agent Interfacetypescript
Multi-LLM Strategy
- Tier 1: Haiku/4o-mini for simple tasks
- Tier 2: Sonnet/4o for moderate complexity
- Tier 3: Opus/o1 for deep reasoning & critical changes
- Fallback chain: Claude → GPT → Gemini → local; prompt adaptation per model and context-window awareness
- Cost model: $1,400/month vs $5,000–$15,000 naive; batch APIs for non-urgent audits; regional routing only if residency mandates
Orchestration
- Backbone: Dagster asset graph with dry-runs, staged rollouts, parallel DAG execution, retries/timeouts/fallbacks, state passing, conditional logic
Command Examples
- Sync governance:
meta-platform sync-governance --domain optimization --all-repos - Compliance across submodules:
meta-platform compliance-check --all-repos - Prompt sync:
meta-platform sync-prompts --all-repos
Cross-References
- Governance tiers ↔ Approval framework; dry-run diffs & staged rollouts feed decisions
- Multi-LLM routing ↔ Agent metrics & cost dashboards; prompt versioning aligns to routing profiles
- Template evolution ↔ Copier migration tests; Phase 2 v1→v2 migrations
- Observability ↔ Grafana dashboards; circuit breakers use signals for self-regulation
Implementation Steps
- Stand up minimal Dagster pipeline: governance scan → dry-run diff → approval decision → canary rollout → metrics emission
- Define LangGraph skeleton for 2 function agents (code-gen, reviewer) and 2 domain agents (Python, Web); instrument metrics
- Create prompts/claude/ routing
[... truncated for PKOS record ...]