@dzhechkov/p-replicator 1.5.8 → 1.5.12

package/templates/.claude/skills/goap-research-ed25519/references/source-evaluation.md

@@ -1,560 +1,59 @@
-# Source Evaluation Reference (Ed25519 Enhanced)
+# Source Evaluation with Ed25519 Provenance
 
-Comprehensive criteria for assessing source credibility with cryptographic trust layers.
+Source evaluation and cryptographic provenance are separate checks.
 
-## Trust Model Overview
+Ed25519 can prove that an active pinned issuer key signed the canonical message and that the signed bytes were not altered. It does not prove that a claim is true. Continue to evaluate methodology, source independence, conflicts of interest, corroboration, and recency.
 
-```
-┌─────────────────────────────────────────────────────────────┐
-│                    TRUST PYRAMID                            │
-├─────────────────────────────────────────────────────────────┤
-│                                                             │
-│                    ┌───────────┐                            │
-│                    │ Level 6   │ ← Cryptographically        │
-│                    │ VERIFIED  │   Signed + Trusted Issuer  │
-│                    └─────┬─────┘                            │
-│                    ┌─────┴─────┐                            │
-│                    │ Level 5   │ ← Primary Sources          │
-│                    │ HIGHEST   │   (traditional)            │
-│                    └─────┬─────┘                            │
-│                ┌─────────┴─────────┐                        │
-│                │     Level 4       │ ← Authoritative         │
-│                │     HIGH          │   Secondary              │
-│                └─────────┬─────────┘                        │
-│            ┌─────────────┴─────────────┐                    │
-│            │        Level 3            │ ← Reputable         │
-│            │        MODERATE           │   General            │
-│            └─────────────┬─────────────┘                    │
-│        ┌─────────────────┴─────────────────┐                │
-│        │           Level 2                 │ ← Community     │
-│        │           LOWER                   │                 │
-│        └─────────────────┬─────────────────┘                │
-│    ┌─────────────────────┴─────────────────────┐            │
-│    │              Level 1                      │ ← Unverified│
-│    │              MINIMAL                      │             │
-│    └───────────────────────────────────────────┘            │
-│                                                             │
-└─────────────────────────────────────────────────────────────┘
-```
-
-## Enhanced Reliability Scoring System
-
-### Level 6: Cryptographically Verified (NEW)
-**Characteristics:**
-- Ed25519 signature verified
-- Issuer in trusted whitelist
-- Content hash matches signature
-- Citation chain intact
-
-**Confidence Score:** 0.95 - 1.00
-
-**Examples:**
-- Academic papers with institutional signatures
-- Government documents with digital signatures
-- Press releases with verified organizational keys
-- Official API responses with authentication
-
-**Verification Requirements:**
-```python
-verified = (
-    signature_valid AND
-    issuer_in_whitelist AND
-    content_hash_matches AND
-    timestamp_valid
-)
-```
-
-**Use for:** Highest-stakes claims, legal evidence, financial data
-
----
-
-### Level 5: Highest Reliability (Traditional)
-**Characteristics:**
-- Primary sources (original research, official documents)
-- Peer-reviewed academic publications
-- Official government/institutional data
-- Direct firsthand accounts from verified parties
-- Original documentation (contracts, records, footage)
-
-**Confidence Score:** 0.85 - 0.95
-
-**Examples:**
-- Published scientific papers in reputable journals
-- Government statistical agencies (census, BLS, WHO)
-- Official company SEC filings, press releases
-- Court documents, legal filings
-- Original research datasets
-
-**Use for:** Foundational facts, statistics, direct evidence
-
----
-
-### Level 4: High Reliability
-**Characteristics:**
-- Authoritative secondary sources
-- Major established publications with editorial oversight
-- Recognized domain experts with verifiable credentials
-- Well-sourced investigative journalism
-- Comprehensive industry reports from established firms
-
-**Confidence Score:** 0.75 - 0.85
-
-**Examples:**
-- Major newspapers (NYT, WSJ, Guardian, Reuters, AP)
-- Academic review articles
-- Industry analysts (Gartner, McKinsey, Forrester)
-- Subject matter experts' published work
-- Technical documentation from official sources
-
-**Use for:** Analysis, expert interpretation, detailed reporting
-
----
-
-### Level 3: Moderate Reliability
-**Characteristics:**
-- Reputable general sources
-- Industry publications and trade journals
-- Established news websites
-- Professional organization publications
-- Verified corporate communications
-
-**Confidence Score:** 0.60 - 0.75
-
-**Examples:**
-- Industry trade publications
-- Regional newspapers
-- Professional association reports
-- Established tech news sites (Ars Technica, Wired)
-- Company blogs (for product information)
-
-**Use for:** Context, industry perspective, supplementary information
+## Levels
 
----
+| Level | Source type | Notes |
+|---|---|---|
+| 5 | Primary documents, official filings, peer-reviewed sources, direct data exports | Strongest non-crypto source tier. |
+| 4 | Established reporting, official company materials, reputable databases | Strong but may have bias or interpretation. |
+| 3 | Industry reports, analyst estimates, expert commentary | Useful with corroboration. |
+| 2 | Blogs, social posts, marketing pages | Use carefully and label limitations. |
+| 1 | Forums, uncited claims, unknown provenance | Low confidence. |
 
-### Level 2: Lower Reliability
-**Characteristics:**
-- Community-generated content with some verification
-- Personal blogs from demonstrated experts
-- User forums with reputation systems
-- Self-published but well-researched content
-- Secondary aggregation sites
+## Crypto Trust Classes
 
-**Confidence Score:** 0.40 - 0.60
+| Trust class | Meaning | Confidence impact |
+|---|---|---|
+| `ISSUER_SIGNED` | Valid signature under active pinned key for the claimed issuer | May reach trusted tier, capped at `0.95`. |
+| `SELF_ATTESTED` | Researcher signed the extracted fact for audit-log tamper-evidence | Capped at `0.60`. |
+| `UNVERIFIED` | Unknown, unsigned, revoked, mismatched, malformed, or invalid signature | `0.0` for crypto provenance. |
 
-**Examples:**
-- Stack Overflow (high-reputation answers)
-- Medium posts by verified professionals
-- Wikipedia (for general context, not citation)
-- Specialized forums (verified contributors)
-- News aggregators
+Invalid signatures are automatic rejection for crypto provenance. They are never a `0.50` multiplier.
 
-**Use for:** Starting points, technical troubleshooting, community consensus
+## Single Confidence Formula
 
----
-
-### Level 1: Minimal Reliability
-**Characteristics:**
-- Unverified sources
-- Anonymous content
-- Social media posts
-- Content with clear commercial bias
-- Rumor or speculation
-
-**Confidence Score:** 0.00 - 0.40
-
-**Examples:**
-- Random social media posts
-- Anonymous forum comments
-- SEO-optimized content farms
-- Press releases (for claims about competitors)
-- Sponsored content disguised as editorial
-
-**Use for:** Leads only—never cite directly without verification
-
----
-
-## Cryptographic Trust Multipliers
-
-When Ed25519 verification is available, apply multipliers:
-
-| Verification Status | Multiplier | Effect |
-|---------------------|------------|--------|
-| Signed + Trusted Issuer | 1.25 | Level 4 → Level 5+ |
-| Signed + Unknown Issuer | 1.10 | Slight boost |
-| Unsigned | 1.00 | No change |
-| Invalid Signature | 0.50 | Major penalty |
-| Revoked Key | 0.00 | Reject source |
-
-### Confidence Calculation Formula
-
-```python
-def calculate_confidence(source):
-    # Base reliability from level
-    base = source.level / 5.0  # 0.2 to 1.0
-    
-    # Verification multiplier
-    if source.signature_verified and source.issuer_trusted:
-        verification_mult = 1.25
-    elif source.signature_verified:
-        verification_mult = 1.10
-    elif source.signature_invalid:
-        verification_mult = 0.50
-    else:
-        verification_mult = 1.00
-    
-    # Recency factor
-    age_days = (now - source.timestamp).days
-    if age_days < 1:
-        recency = 1.00
-    elif age_days < 7:
-        recency = 0.95
-    elif age_days < 30:
-        recency = 0.85
-    elif age_days < 365:
-        recency = 0.70
-    else:
-        recency = 0.50
-    
-    # Cap at 1.0
-    return min(1.0, base * verification_mult * recency)
 ```
-
----
-
-## Source Evaluation Checklist (Enhanced)
-
-### Authority Assessment
-- [ ] Author/organization identifiable and credentialed
-- [ ] Publication has editorial standards
-- [ ] Expert credentials relevant to topic
-- [ ] Track record of accuracy
-- [ ] Institutional backing or peer review
-- [ ] **Ed25519 public key registered** (NEW)
-- [ ] **Issuer in trusted whitelist** (NEW)
-
-### Cryptographic Assessment (NEW)
-- [ ] Content signed with Ed25519
-- [ ] Signature verification passes
-- [ ] Public key retrievable from trusted source
-- [ ] Issuer domain matches claimed origin
-- [ ] No key revocation notices
-- [ ] Timestamp within acceptable range
-
-### Currency Assessment
-- [ ] Publication date clearly stated
-- [ ] Information current for topic type
-- [ ] No signs of outdated information
-- [ ] Updates/corrections noted if applicable
-- [ ] **Signature timestamp recent** (NEW)
-
-### Coverage Assessment
-- [ ] Scope appropriate to claims made
-- [ ] Evidence provided for assertions
-- [ ] Methodology explained for research
-- [ ] Limitations acknowledged
-- [ ] Multiple perspectives considered
-
-### Objectivity Assessment
-- [ ] Potential biases identified and manageable
-- [ ] Funding/sponsorship disclosed
-- [ ] Language neutral (not promotional)
-- [ ] Counterarguments addressed
-- [ ] Commercial interests declared
-
-### Accuracy Assessment
-- [ ] Facts verifiable through other sources
-- [ ] Statistics include methodology
-- [ ] Quotes attributed and checkable
-- [ ] No obvious errors detected
-- [ ] Sources cited properly
-- [ ] **Content hash matches signature** (NEW)
-
----
-
-## Trusted Issuer Categories
-
-### Tier 1: Government & Regulatory
-**Trust Level:** Highest (auto-Level 5+)
-```yaml
-tier1_issuers:
-  - "*.gov"
-  - "*.gov.uk"
-  - "*.gov.au"
-  - "europa.eu"
-  - "who.int"
-  - "worldbank.org"
-  - "imf.org"
+if trust_class == UNVERIFIED:
+    confidence = 0.0
+elif trust_class == SELF_ATTESTED:
+    confidence = min(0.60, source_level / 5 * recency_factor)
+elif trust_class == ISSUER_SIGNED:
+    confidence = min(0.95, source_level / 5 * recency_factor)
 ```
 
-### Tier 2: Academic & Research
-**Trust Level:** Very High (Level 5)
-```yaml
-tier2_issuers:
-  - "arxiv.org"
-  - "nature.com"
-  - "science.org"
-  - "sciencedirect.com"
-  - "springer.com"
-  - "ieee.org"
-  - "acm.org"
-  - "pubmed.gov"
-  - "*.edu"
-```
+Recency factors are domain-specific. A common default is `1.0` for current sources, `0.9` for recent sources, `0.8` for older but still relevant sources, and lower when the domain changes quickly.
 
-### Tier 3: Major News & Wire Services
-**Trust Level:** High (Level 4+)
-```yaml
-tier3_issuers:
-  - "reuters.com"
-  - "ap.org"
-  - "afp.com"
-  - "bbc.com"
-  - "nytimes.com"
-  - "wsj.com"
-  - "washingtonpost.com"
-  - "economist.com"
-  - "ft.com"
-```
+## Mode Behavior
 
-### Tier 4: Industry & Professional
-**Trust Level:** Moderate-High (Level 3-4)
-```yaml
-tier4_issuers:
-  - "gartner.com"
-  - "mckinsey.com"
-  - "forrester.com"
-  - "deloitte.com"
-  - "pwc.com"
-  - "kpmg.com"
-```
-
----
-
-## Domain-Specific Evaluation (Enhanced)
-
-### Scientific/Medical Information
-**Require:** 
-- Peer-reviewed sources, established journals
-- **Ed25519 signature from institutional key**
-
-**Watch for:** 
-- Predatory journals
-- Non-replicated studies
-- Press releases overstating findings
-- **Signatures from unrecognized institutions**
-
-**Verify:** 
-- Sample sizes, methodology, conflicts of interest
-- Replication status
-- **Institutional key registry**
-
-**Minimum Requirements:**
-```
-Mode: strict (0.95)
-Sources: ≥2 Level 5+
-Signatures: Required from Tier 2 issuers
-```
-
-### Financial/Business Information
-**Require:** 
-- Official filings, audited statements
-- **SEC/regulatory body signatures where applicable**
-
-**Watch for:** 
-- Promotional content
-- Undisclosed sponsored research
-- **Forged corporate signatures**
-
-**Verify:** 
-- Numbers against multiple independent sources
-- Check for corrections
-- **Verify corporate key against official registry**
-
-### Technology Information
-**Require:** 
-- Official documentation
-- Benchmarks with methodology
-- **Signed releases from official project keys**
-
-**Watch for:** 
-- Marketing claims
-- Outdated information (rapid change)
-- **Compromised project keys**
-
-**Verify:** 
-- Version numbers
-- Test conditions
-- Community feedback
-- **Project key from official source (GitHub, website)**
-
-### Legal Information
-**Require:** 
-- Primary legal sources
-- Licensed attorney commentary
-- **Court/government signatures**
-
-**Watch for:** 
-- Jurisdiction-specific variations
-- Outdated precedents
-- **Unofficial legal advice**
-
-**Verify:** 
-- Current status of laws/cases
-- Jurisdictional applicability
-- **Government issuer keys**
-
-### Political/News Information
-**Require:** 
-- Multiple independent sources
-- Primary documentation
-- **≥3 Tier 3 issuer signatures**
-
-**Watch for:** 
-- Partisan framing
-- Selective reporting
-- Anonymous sourcing
-- **Single-source "scoops" without verification**
-
-**Verify:** 
-- Cross-reference across ideologically diverse outlets
-- **Multiple independent signatures**
-
----
-
-## Bias Recognition Matrix (Enhanced)
+| Mode | Unsigned or invalid claims |
+|---|---|
+| `development` | Allowed only when labeled. |
+| `moderate` | Allowed with lower confidence and explicit caveats. |
+| `strict` | Rejected. |
+| `paranoid` | Rejected. |
 
-| Bias Type | Indicators | Mitigation | Verification Help |
-|-----------|------------|------------|-------------------|
-| Commercial | Promotes product/service, lacks criticism | Seek independent reviews | Require non-commercial issuer |
-| Political | One-sided framing, loaded language | Cross-reference opposing viewpoints | Multi-perspective signatures |
-| Confirmation | Only supports pre-existing narrative | Actively seek disconfirming evidence | Signed dissenting sources |
-| Recency | Overweights recent events | Include historical context | Check timestamp validity |
-| Survivorship | Only successful examples shown | Look for failure cases | Academic peer-review sigs |
-| Selection | Cherry-picked data/examples | Seek comprehensive datasets | Raw data signatures |
-| **Forgery** (NEW) | Claimed sources don't exist | **Verify signatures** | Ed25519 verification |
+The planner enforces strict/paranoid rejection by discarding completed plans that still contain unsigned claims.
 
----
+## Practical Checklist
 
-## Verification Requirements by Claim Type (Enhanced)
-
-### Hard Facts (dates, numbers, names)
-**Traditional:**
-- Minimum: 1 Level 4+ source
-- Preferred: Primary source documentation
-- Red flag: Only Level 2 or below available
-
-**With Ed25519:**
-- Minimum: 1 Level 4+ source OR 1 signed Level 3 source
-- Preferred: Signed primary source
-- Auto-accept: Level 6 (cryptographically verified)
-
-### Expert Opinions
-**Traditional:**
-- Minimum: Attributed to named, credentialed expert
-- Preferred: Multiple experts agreeing
-- Red flag: Anonymous or unverifiable credentials
-
-**With Ed25519:**
-- Minimum: Signed by expert's verified key
-- Preferred: Institutional countersignature
-- Auto-accept: Academic institution signature
-
-### Statistics
-**Traditional:**
-- Minimum: Methodology disclosed, Level 4+ source
-- Preferred: Original study/dataset accessible
-- Red flag: No methodology, round numbers without context
-
-**With Ed25519:**
-- Minimum: Signed by research institution
-- Preferred: Signed raw data + methodology
-- Required for strict mode: Institutional signature
-
-### Predictions/Forecasts
-- Label confidence explicitly
-- Note track record of forecaster
-- Present range of expert opinions
-- **Never auto-accept signed predictions as facts**
-
-### Controversial Claims
-**Traditional:**
-- Minimum: 3+ independent Level 3+ sources
-- Preferred: Primary evidence available
-- Required: Opposing viewpoints documented
-
-**With Ed25519:**
-- Minimum: 3+ independently signed sources
-- Preferred: Signed primary evidence + 2 corroborations
-- Required: Signed opposing viewpoint included
-
----
-
-## Anti-Hallucination Verification Matrix
-
-| Claim Type | Unsigned OK? | Signature Required? | Strict Mode |
-|------------|--------------|---------------------|-------------|
-| Historical fact | Yes (Level 4+) | Preferred | Level 5+ |
-| Current event | Level 4+ only | Required | Tier 3 sig |
-| Scientific claim | Level 5+ only | Required | Tier 2 sig |
-| Financial data | No | Required | Regulatory sig |
-| Medical advice | No | Required | Institutional sig |
-| Legal statement | No | Required | Gov/court sig |
-| Expert quote | Level 4+ | Preferred | Expert key |
-| Statistics | Level 5+ | Required | Methodology sig |
-
----
-
-## Source Rejection Criteria
-
-Automatically reject sources with:
-- Invalid Ed25519 signature
-- Revoked issuer key
-- Timestamp >1 year old for current events
-- Mismatched content hash
-- Unknown issuer in strict mode
-- Multiple failed verification attempts
-- Blacklisted domain
-
----
-
-## Confidence Thresholds by Research Mode
-
-| Mode | Threshold | Unsigned Sources | Signature Required |
-|------|-----------|------------------|-------------------|
-| Development | 0.75 | Allowed (Level 3+) | Optional |
-| Moderate | 0.85 | Limited (Level 4+) | Preferred |
-| Strict | 0.95 | Rare (Level 5 only) | Required |
-| Paranoid | 0.99 | Never | Always (Tier 1-2) |
-
----
-
-## Appendix: Quick Reference Card
-
-```
-┌────────────────────────────────────────────────────────┐
-│          SOURCE EVALUATION QUICK REFERENCE             │
-├────────────────────────────────────────────────────────┤
-│ ALWAYS VERIFY:                                         │
-│   ✓ Ed25519 signature (if available)                  │
-│   ✓ Issuer against whitelist                          │
-│   ✓ Content hash integrity                            │
-│   ✓ Timestamp validity                                │
-│                                                        │
-│ CONFIDENCE FORMULA:                                    │
-│   confidence = level/5 × verification × recency        │
-│                                                        │
-│ MINIMUM REQUIREMENTS:                                  │
-│   Facts: 1 signed source OR 2 Level 4+ unsigned       │
-│   Stats: Signed methodology OR Level 5+ with method   │
-│   Quotes: Signed OR Level 4+ with attribution         │
-│                                                        │
-│ RED FLAGS:                                             │
-│   ✗ Invalid/missing signature in strict mode          │
-│   ✗ Unknown issuer for sensitive claims               │
-│   ✗ Single source for controversial claims            │
-│   ✗ No methodology for statistics                     │
-│                                                        │
-│ ANTI-HALLUCINATION RULE:                              │
-│   "If you can't cite it, don't claim it"              │
-└────────────────────────────────────────────────────────┘
-```
+- Identify the source URL for each claim.
+- Record source level and recency.
+- If using issuer-grade crypto, verify against a pinned active issuer key.
+- Reject unknown, revoked, mismatched, malformed, or invalid signatures.
+- Treat self-attestation as audit evidence only.
+- Cross-check important claims through independent sources.
+- Report limitations and unresolved uncertainty explicitly.