athena-agent 10.0.0__tar.gz

athena_agent-10.0.0/.agent/PROMPT_LIBRARY.md

@@ -0,0 +1,306 @@
+# Athena Prompt Library
+
+> **Purpose**: Curated, reusable system prompts for steering AI behavior.  
+> **Format**: Stealable meta-prompts (per `Core_Identity.md` standard).  
+> **Last Updated**: 2026-07-22
+
+---
+
+## Index
+
+| Category | Count |
+|----------|-------|
+| [Strategy](#strategy) | 4 |
+| [Analysis](#analysis) | 3 |
+| [Coding](#coding) | 3 |
+| [Writing](#writing) | 2 |
+| [Research](#research) | 2 |
+
+---
+
+## Strategy
+
+### The Pre-Mortem
+
+**Tags**: `decision` `risk` `planning`  
+**Use Case**: Before committing to any major decision, force failure-mode thinking.
+
+```
+You are a strategic analyst. Before I commit to this decision, conduct a pre-mortem:
+
+1. Assume the decision has FAILED catastrophically 1 year from now.
+2. List the 5 most likely reasons for failure.
+3. For each reason, identify:
+   - Warning signs I should watch for
+   - Preventive actions I can take NOW
+4. Rate overall decision robustness (1-10) after this analysis.
+
+Decision: [INSERT]
+```
+
+---
+
+### The Counter-Argument Generator
+
+**Tags**: `decision` `adversarial` `debate`  
+**Use Case**: Steel-man the opposing position before finalizing a stance.
+
+```
+You are an adversarial debate partner. Your job is to ATTACK my position as strongly as possible.
+
+1. Identify the 3 strongest arguments AGAINST my position.
+2. For each argument, provide evidence or logic that supports it.
+3. Identify the weakest assumption in my reasoning.
+4. If you were betting against me, what would you bet on?
+
+My position: [INSERT]
+```
+
+---
+
+### The Opportunity Cost Calculator
+
+**Tags**: `decision` `economics` `tradeoffs`  
+**Use Case**: Force explicit consideration of what you're giving up.
+
+```
+You are an economics advisor. For every hour/dollar I spend on Option A, I'm NOT spending it elsewhere.
+
+1. List the top 3 alternative uses for this resource (time/money/attention).
+2. Estimate the expected value of each alternative.
+3. Calculate the TRUE cost of Option A (direct cost + best alternative foregone).
+4. Verdict: Is Option A still worth it after accounting for opportunity cost?
+
+Option A: [INSERT]
+Resources at stake: [INSERT]
+```
+
+---
+
+### The Second-Order Effects Scanner
+
+**Tags**: `strategy` `systems` `consequences`  
+**Use Case**: Anticipate downstream effects of a decision.
+
+```
+You are a systems thinker. Analyze the ripple effects of this decision:
+
+1. **First-order effects**: What happens immediately/directly?
+2. **Second-order effects**: What happens as a RESPONSE to the first-order effects?
+3. **Third-order effects**: What feedback loops or cascades might emerge?
+4. **Unintended consequences**: What could go wrong that isn't obvious?
+
+Decision/Action: [INSERT]
+```
+
+---
+
+## Analysis
+
+### The MECE Breakdown
+
+**Tags**: `analysis` `structure` `decomposition`  
+**Use Case**: Ensure exhaustive, non-overlapping analysis of a problem.
+
+```
+You are a management consultant. Break down this problem using MECE (Mutually Exclusive, Collectively Exhaustive) principles:
+
+1. Identify all major categories of the problem (no overlaps).
+2. Ensure categories cover 100% of the problem space.
+3. For each category, list 2-3 sub-components.
+4. Highlight which category has the highest leverage for solving the problem.
+
+Problem: [INSERT]
+```
+
+---
+
+### The Assumption Auditor
+
+**Tags**: `analysis` `critical-thinking` `validation`  
+**Use Case**: Surface hidden assumptions before they cause problems.
+
+```
+You are a critical thinking coach. Audit my reasoning for hidden assumptions:
+
+1. List all implicit assumptions in my statement/plan.
+2. For each assumption, rate:
+   - Validity (1-10): How likely is this assumption true?
+   - Impact (1-10): If wrong, how badly does it break the plan?
+3. Flag any "load-bearing" assumptions (high impact + uncertain validity).
+4. Suggest validation methods for the top 3 riskiest assumptions.
+
+My statement/plan: [INSERT]
+```
+
+---
+
+### The Inversion Prompt
+
+**Tags**: `analysis` `problem-solving` `inversion`  
+**Use Case**: Solve a problem by first solving its opposite.
+
+```
+You are a problem-solving expert using Charlie Munger's inversion technique.
+
+Instead of asking "How do I achieve X?", let's ask "How would I GUARANTEE failure at X?"
+
+1. List 5 surefire ways to FAIL at this goal.
+2. Invert each failure mode into a success principle.
+3. Which inverted principle is most actionable RIGHT NOW?
+
+Goal: [INSERT]
+```
+
+---
+
+## Coding
+
+### The Code Explainer
+
+**Tags**: `coding` `learning` `documentation`  
+**Use Case**: Get educational breakdown of unfamiliar code.
+
+```
+You are a senior developer and educator. Explain this code to me:
+
+1. **High-level purpose**: What does this code accomplish in one sentence?
+2. **Step-by-step breakdown**: Walk through each section/function.
+3. **Key concepts**: What programming patterns or concepts does this use?
+4. **Gotchas**: What's non-obvious or easy to get wrong?
+5. **Improvement suggestions**: How would a senior dev refactor this?
+
+Code: [INSERT]
+```
+
+---
+
+### The Debug Partner
+
+**Tags**: `coding` `debugging` `troubleshooting`  
+**Use Case**: Systematic debugging when stuck on an issue.
+
+```
+You are a debugging expert. Help me systematically diagnose this issue:
+
+1. **Symptom summary**: Restate the problem in your own words.
+2. **Hypothesis generation**: List 3-5 possible causes, ranked by likelihood.
+3. **Diagnostic steps**: For each hypothesis, what test would confirm/eliminate it?
+4. **Quick wins**: What's the fastest thing to try first?
+
+Error/Issue: [INSERT]
+Context: [INSERT]
+```
+
+---
+
+### The Code Review Prompt
+
+**Tags**: `coding` `quality` `review`  
+**Use Case**: Comprehensive code review checklist.
+
+```
+You are a senior code reviewer. Review this code for:
+
+1. **Correctness**: Does it do what it's supposed to do?
+2. **Readability**: Is it easy to understand? Naming? Comments?
+3. **Performance**: Any obvious inefficiencies or N+1 queries?
+4. **Security**: Any vulnerabilities (injection, auth, data exposure)?
+5. **Edge cases**: What inputs would break this?
+6. **Testability**: Is this easy to unit test?
+
+Provide specific line-level feedback with suggested fixes.
+
+Code: [INSERT]
+```
+
+---
+
+## Writing
+
+### The Writing Improver
+
+**Tags**: `writing` `editing` `polish`  
+**Use Case**: Upgrade draft writing while preserving voice.
+
+```
+You are a professional editor. Improve this writing while preserving my voice:
+
+1. Fix grammar, punctuation, and typos.
+2. Improve clarity and flow.
+3. Cut unnecessary words (aim for 20% reduction).
+4. Strengthen weak verbs and vague language.
+5. Show your changes as tracked edits with brief explanations.
+
+Draft: [INSERT]
+```
+
+---
+
+### The Hook Generator
+
+**Tags**: `writing` `marketing` `copywriting`  
+**Use Case**: Generate attention-grabbing opening lines.
+
+```
+You are a world-class copywriter. Generate 5 hook options for this piece:
+
+Context: [What is this content about?]
+Audience: [Who is reading this?]
+Goal: [What action do I want them to take?]
+
+For each hook, specify which psychological trigger it uses (curiosity, fear, desire, urgency, etc.).
+```
+
+---
+
+## Research
+
+### The Deep Dive Prompt
+
+**Tags**: `research` `learning` `comprehensive`  
+**Use Case**: Exhaustive exploration of an unfamiliar topic.
+
+```
+You are a research expert. Conduct a deep dive on this topic:
+
+1. **Overview**: What is this? Why does it matter?
+2. **History**: Key milestones and how we got here.
+3. **Current state**: What's happening NOW?
+4. **Key players**: Who are the important people/companies/institutions?
+5. **Debates**: What are the open questions or controversies?
+6. **Resources**: What should I read/watch next for deeper understanding?
+
+Topic: [INSERT]
+```
+
+---
+
+### The Source Validator
+
+**Tags**: `research` `credibility` `fact-checking`  
+**Use Case**: Evaluate reliability of a source or claim.
+
+```
+You are a fact-checking expert. Evaluate this source/claim:
+
+1. **Source credibility**: Who published this? What's their track record?
+2. **Evidence quality**: What evidence supports the claim? Primary vs secondary?
+3. **Bias check**: What incentives might distort this source's perspective?
+4. **Corroboration**: Can this be verified from independent sources?
+5. **Verdict**: Reliability rating (1-10) with justification.
+
+Source/Claim: [INSERT]
+```
+
+---
+
+## Changelog
+
+| Date | Change |
+|------|--------|
+| 2026-01-11 | Initial creation. 14 seed prompts across 5 categories. |
+
+---
+
+# prompt-library #stealable #meta-prompts

athena_agent-10.0.0/.agent/config/CAPS.json

@@ -0,0 +1,86 @@
+{
+  "_comment": "Canonical inventory counts for Project Athena. Single source of truth for pre-commit caps, audit asserts, and AGENTS.md regeneration. Update via .agent/scripts/sync_agents_md.py (runs automatically: pre-commit Gate 4 + shutdown.py Phase 1.5), not by hand.",
+  "_generated": "2026-09-28",
+  "_generator": "S411 sync_agents_md.py automation (TD-028)",
+  "version": {
+    "framework": "v8.2-stable",
+    "system": "v10.0.0"
+  },
+  "protocols": {
+    "total": 456,
+    "active": 422,
+    "archived": 34,
+    "categories": 26,
+    "cap": "uncapped",
+    "cap_policy": "Protocols are uncapped as of v9.8.0. The constraint is on BOOT LOAD (attention budget), not on total count. Protocol 530 (Conditional Skill Loading) + Progressive Disclosure ensure only relevant protocols enter context per session. Archived protocols don't affect boot. TD-019 is now structurally resolved."
+  },
+  "skills": {
+    "active": 43,
+    "archived": 17,
+    "conditional_count": 42,
+    "always_on_count": 10
+  },
+  "workflows": {
+    "root": 55,
+    "domain": 20,
+    "total": 75,
+    "cap": "uncapped",
+    "cap_policy": "Workflows are uncapped as of v9.8.0. Domain workflows are dormant by default (_domain/ tier) and only activate on query match. The constraint is attention budget, not inventory size. TD-017 is now structurally resolved."
+  },
+  "canonical": {
+    "total_entries": 199,
+    "cap": "uncapped",
+    "cap_policy": "Entry count is uncapped as of 2026-05-07 (GTO alignment with protocols/workflows). The constraint is boot-load attention budget (~35K token ceiling / ~140KB), not entry count. Staleness thresholds (3mo/6mo) + quarterly /audit-canonical are the quality gates. Tiered loading ensures only relevant entries enter context per session.",
+    "attention_budget_ceiling_tokens": 35000,
+    "attention_budget_ceiling_bytes": 140000,
+    "current_bytes": 57458,
+    "current_utilization_pct": 41,
+    "tier_1_always_boot": 40,
+    "tier_2_domain_triggered": 156,
+    "tier_3_on_demand": 3,
+    "boot_savings_pct": 69,
+    "tier_map_path": ".agent/telemetry/tier_map.json"
+  },
+  "context_files": {
+    "glossary": ".context/glossary.md",
+    "connectors": ".agent/CONNECTORS.md",
+    "_note": "glossary.md and CONNECTORS.md added 2026-05-24 (Anthropic steal)"
+  },
+  "telemetry": {
+    "retrieval_log": ".agent/telemetry/retrieval_log.jsonl",
+    "retrieval_audit_script": ".agent/scripts/retrieval_audit.py",
+    "canonical_tier_analysis": ".agent/scripts/canonical_tier_analysis.py"
+  },
+  "scripts": 285,
+  "memories": {
+    "total_files": 5043,
+    "path": ".context/memories/"
+  },
+  "paths": {
+    "active_protocols": ".agent/skills/protocols/",
+    "archived_protocols": ".agent/skills/protocols/archive/",
+    "active_skills": ".agent/skills/",
+    "archived_skills": ".agent/archive_skills/",
+    "root_workflows": ".agent/workflows/",
+    "domain_workflows": ".agent/workflows/_domain/",
+    "telemetry": ".agent/telemetry/"
+  },
+  "recount_rules": {
+    "_note": "All rules exclude README.md so archive marker files don't inflate counts.",
+    "protocols_active": "find .agent/skills/protocols -type f -name '*.md' -not -path '*/archive/*' -not -name 'README.md' | wc -l",
+    "protocols_total": "find .agent/skills/protocols -type f -name '*.md' -not -name 'README.md' | wc -l",
+    "protocols_archived": "find .agent/skills/protocols/archive -type f -name '*.md' -not -name 'README.md' | wc -l",
+    "protocols_categories": "find .agent/skills/protocols -mindepth 1 -maxdepth 1 -type d -not -name archive | wc -l",
+    "skills_active": "ls -d .agent/skills/*/SKILL.md 2>/dev/null | wc -l",
+    "skills_archived": "ls -d .agent/archive_skills/*/ | wc -l",
+    "workflows_root": "find .agent/workflows -maxdepth 1 -type f -name '*.md' | wc -l",
+    "workflows_domain": "find .agent/workflows/_domain -type f -name '*.md' | wc -l",
+    "conditional_skills": "grep -l 'context_trigger' .agent/skills/*/SKILL.md | wc -l",
+    "scripts": "find .agent/scripts -type f | wc -l"
+  },
+  "case_studies": {
+    "total_files": 503,
+    "uppercase_convention": 497,
+    "note": "497 CS-*.md + 6 legacy cs-*.md (now renamed). 7 indexed without files, 3 unindexed. See CASE_STUDY_INDEX.md."
+  }
+}

athena_agent-10.0.0/.agent/config/README.md

@@ -0,0 +1,41 @@
+# Athena Agent Manifest
+
+This is the **machine-readable source of truth** for the Athena Sovereign Agent architecture.
+
+## What It Does
+
+The manifest defines:
+
+- **Session Protocol**: Boot (`/start`) and commit (`/end`) lifecycle hooks
+- **Retrieval Pipeline**: RRF Fusion with configurable source weights
+- **Orchestration**: ReAct loop with escalation to Tree/Graph search
+- **Trilateral Feedback**: Cross-model disagreement detection
+- **Budgets**: Token, tool call, and cost limits per session
+
+## Integration
+
+```python
+from scripts.core.config_loader import ManifestLoader
+
+# Load config
+budgets = ManifestLoader.get_budget_config()
+print(f"Token budget: {budgets.token_budget:,}")
+
+# Get retrieval weights
+retrieval = ManifestLoader.get_retrieval_config()
+print(f"RRF k: {retrieval.rrf_k}")
+```
+
+## Files
+
+| File | Purpose |
+|------|---------|
+| `.agent/config/athena.agent.manifest.json` | Configuration kernel |
+| `scripts/core/retrieval/pipeline.py` | RRF Fusion implementation |
+| `scripts/core/reflection.py` | Reflexion pattern |
+| `scripts/core/auditor.py` | Trilateral feedback |
+
+## Documentation
+
+- [Architecture Overview](../../docs/ARCHITECTURE.md)
+- [Protocol 415: Exoskeleton](../../docs/protocols/415-exoskeleton-architecture.md)

athena_agent-10.0.0/.agent/config/ag_middleware.json

@@ -0,0 +1,7 @@
+{
+  "agent": "athena",
+  "workspace_root": ".",
+  "mcp_server": "python3 src/athena/mcp_server.py",
+  "pre_action_hook": "python3 -c 'from athena.core.gate import AgentGate; AgentGate(\".\").intercept_tool(...)'",
+  "post_action_hook": null
+}

athena_agent-10.0.0/.agent/config/gate_config.json

@@ -0,0 +1,26 @@
+{
+  "version": "10.0.0",
+  "universal_gate_enabled": true,
+  "meta_awareness_structural_classes": [
+    "T1-INBOUND",
+    "T2-OUTBOUND",
+    "T3-VERDICT",
+    "T4-RESOURCE",
+    "T5-FELT"
+  ],
+  "anti_sycophancy_kernel": true,
+  "ruin_check_mode": "structured",
+  "ide_specific_hooks": {
+    "claude_code": {
+      "enabled": true,
+      "hook_path": ".claude/settings.json",
+      "hook_command": "python3 \"$CLAUDE_PROJECT_DIR/.agent/scripts/hook_meta_awareness_gate.py\""
+    },
+    "antigravity": {
+      "enabled": true,
+      "middleware_config_path": "~/.gemini/antigravity/mcp_config.json",
+      "mcp_integration": true,
+      "agent_manifest_override": true
+    }
+  }
+}

athena_agent-10.0.0/.agent/config/harness_scorecard.yaml

@@ -0,0 +1,210 @@
+# .agent/config/harness_scorecard.yaml
+# Single source of truth for Athena Harness Capability & Enforcement scoring.
+version: 1
+updated: "2026-08-24"
+
+pillars:
+  P1:
+    name: "Harness Engineering & Deterministic Verification"
+    weight: 22
+    sub_variables:
+      1.1:
+        name: "Session lifecycle hooks"
+        w: 0.12
+        C: 0.90
+        probe: "session_lifecycle"
+      1.2:
+        name: "Input-side gates"
+        w: 0.12
+        C: 0.95
+        probe: "input_gates"
+      1.3:
+        name: "Output-side gate before user delivery"
+        w: 0.20
+        C: 0.85
+        probe: "output_stop_gate"
+      1.4:
+        name: "Repo gates at commit"
+        w: 0.12
+        C: 0.95
+        probe: "git_hooks_path"
+      1.5:
+        name: "CI gates"
+        w: 0.12
+        C: 0.85
+        probe: "ci_hard_fail"
+      1.6:
+        name: "PreToolUse secret/ruin interception"
+        w: 0.10
+        C: 0.85
+        probe: "pre_tool_secrets"
+      1.7:
+        name: "Loop / budget / no-progress middleware"
+        w: 0.10
+        C: 0.40
+        probe: "loop_middleware"
+      1.8:
+        name: "Sandboxed execution & worktree isolation"
+        w: 0.07
+        C: 0.60
+        probe: "worktree_isolation"
+      1.9:
+        name: "Oracle-tamper protection"
+        w: 0.05
+        C: 0.60
+        probe: "oracle_protection"
+
+  P2:
+    name: "Context Engineering & Memory"
+    weight: 18
+    sub_variables:
+      2.1:
+        name: "Tiered canonical loading"
+        w: 0.18
+        C: 0.90
+        probe: "tiered_loading"
+      2.2:
+        name: "Progressive skill disclosure"
+        w: 0.15
+        C: 0.85
+        probe: "skill_triggers"
+      2.3:
+        name: "Durable session state"
+        w: 0.15
+        C: 0.85
+        probe: "durable_session"
+      2.4:
+        name: "Retrieval quality runtime"
+        w: 0.20
+        C: 0.85
+        probe: "reranker_runtime"
+      2.5:
+        name: "Structured cross-window handoff"
+        w: 0.12
+        C: 0.60
+        probe: "structured_handoff"
+      2.6:
+        name: "Compaction & clearing discipline"
+        w: 0.10
+        C: 0.80
+        probe: "compaction_tools"
+      2.7:
+        name: "Retrieval regression acted on"
+        w: 0.10
+        C: 0.90
+        probe: "eval_regression_actuator"
+
+  P3:
+    name: "Spec-Driven Development"
+    weight: 15
+    sub_variables:
+      3.1:
+        name: "Plan/execute separation"
+        w: 0.30
+        C: 0.90
+        probe: "plan_execute_separation"
+      3.2:
+        name: "Machine-checkable definition of done"
+        w: 0.30
+        C: 0.60
+        probe: "machine_dod"
+      3.3:
+        name: "Interrogation ritual"
+        w: 0.20
+        C: 0.85
+        probe: "interrogation_ritual"
+      3.4:
+        name: "Spec-artifact drift check"
+        w: 0.20
+        C: 0.40
+        probe: "spec_drift"
+
+  P4:
+    name: "Automated Evaluation, Oracles & Critique"
+    weight: 20
+    sub_variables:
+      4.1:
+        name: "Deterministic test suite for own code"
+        w: 0.18
+        C: 0.85
+        probe: "test_suite"
+      4.2:
+        name: "Golden retrieval eval + baseline compare"
+        w: 0.18
+        C: 0.90
+        probe: "golden_eval_recent"
+      4.3:
+        name: "Output-quality oracle on deliverable"
+        w: 0.22
+        C: 0.70
+        probe: "output_quality_oracle"
+      4.4:
+        name: "Independent critic in fresh context"
+        w: 0.15
+        C: 0.50
+        probe: "independent_critic"
+      4.5:
+        name: "Behavioural golden set"
+        w: 0.12
+        C: 0.60
+        probe: "behavioral_golden"
+      4.6:
+        name: "Scheduled runs & regression alerting"
+        w: 0.15
+        C: 0.40
+        probe: "scheduled_eval"
+
+  P5:
+    name: "Tool & API Ergonomics"
+    weight: 12
+    sub_variables:
+      5.1:
+        name: "Right-sized, searchable tool surface"
+        w: 0.25
+        C: 0.80
+        probe: "tool_search"
+      5.2:
+        name: "Actionable error payloads"
+        w: 0.25
+        C: 0.50
+        probe: "error_payloads"
+      5.3:
+        name: "Token-bounded responses"
+        w: 0.20
+        C: 0.60
+        probe: "token_bounds"
+      5.4:
+        name: "Fail-loud contract"
+        w: 0.30
+        C: 0.50
+        probe: "fail_loud"
+
+  P6:
+    name: "Skill Compilation & Compounded Learning"
+    weight: 13
+    sub_variables:
+      6.1:
+        name: "Curated, indexed, archived skill library"
+        w: 0.20
+        C: 0.90
+        probe: "skill_library"
+      6.2:
+        name: "Automatic solved-to-skill compilation"
+        w: 0.25
+        C: 0.50
+        probe: "skill_compiler_trigger"
+      6.3:
+        name: "Failure-to-lesson mining from traces"
+        w: 0.25
+        C: 0.85
+        probe: "reflexion_harvester"
+      6.4:
+        name: "Delta-merge discipline"
+        w: 0.15
+        C: 0.30
+        probe: "delta_merge"
+      6.5:
+        name: "Anti-bloat hygiene"
+        w: 0.15
+        C: 0.80
+        probe: "anti_bloat_hygiene"