0d5eeeae0c
Checkpoint: File edits
...
Automatic checkpoint created by Claude Code
- Branch: claude/guidance-control-plane-uhmR3
- Timestamp: 2026-02-02T03:52:19Z
- Changes: 1 file(s)
🤖 Generated with [Claude Code](https://claude.com/claude-code )
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com >
2026-02-02 03:52:19 +00:00
89342d8ae8
Checkpoint: File edits
...
Automatic checkpoint created by Claude Code
- Branch: claude/guidance-control-plane-uhmR3
- Timestamp: 2026-02-02T03:52:12Z
- Changes: 1 file(s)
🤖 Generated with [Claude Code](https://claude.com/claude-code )
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com >
2026-02-02 03:52:12 +00:00
15eaaf8022
Checkpoint: File edits
...
Automatic checkpoint created by Claude Code
- Branch: claude/guidance-control-plane-uhmR3
- Timestamp: 2026-02-02T03:52:09Z
- Changes: 1 file(s)
🤖 Generated with [Claude Code](https://claude.com/claude-code )
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com >
2026-02-02 03:52:09 +00:00
3e5f92ef8b
security: eliminate hardcoded HMAC keys, fix timing attacks, remove command injection
...
- Remove hardcoded DEFAULT_SIGNING_KEY from proof.ts, artifacts.ts, evolution.ts
Constructors now require explicit signingKey (fail-closed)
- Add timingSafeEqual() XOR-based constant-time comparison to proof.ts,
artifacts.ts, authority.ts (prevents HMAC timing side-channels)
- Fix command injection in headless.ts: replace execFile('sh', ['-c', cmd])
with direct arg-array execFile('claude', ['-p', prompt, ...])
- Fix command injection in analyzer.ts: same shell-free pattern
- Remove dead code: unused toLowerCase() in analyzer.ts extractMetrics
and scoreCompleteness
- Hoist gates.ts GATE_DECISION_SEVERITY to module-level constant
- Add ConformanceRunner signingKey parameter for explicit key propagation
- Update tests: 3 new "should throw without key" assertions
- Add flight build report (docs/reports/flight-build-report.md)
1,331 tests passing.
https://claude.ai/code/session_01HLi9WmfvFgkDLJxLKGAUTf
2026-02-02 01:51:26 +00:00
671831dacd
feat: add A/B benchmark harness with 20 tasks, gate simulation, and composite scoring
...
Implements the Measurement Plan: Config A (no control plane) vs Config B
(Phase 1 guidance) across 20 representative tasks from Claude Flow repo
history spanning 7 task classes (bug-fix, feature, refactor, security,
deployment, test, performance).
KPIs: success rate, tool calls, token spend, violations, human
interventions. Composite score formula:
score = success_rate - 0.1*cost - 0.2*violations - 0.1*interventions
Category shift detection: B beats A by ≥0.2 on composite across ≥3 task
classes. Gate simulation detects destructive commands, hardcoded secrets,
force push, unsafe types, skipped hooks, missing tests, policy violations.
1,328 tests passing across 26 files (38 new AB benchmark tests).
https://claude.ai/code/session_01HLi9WmfvFgkDLJxLKGAUTf
2026-02-01 22:22:17 +00:00
2810c8bdc4
docs: update README and API reference with generators, analyzer, and validation suite
...
Add documentation for all features built in recent sessions:
- Generators module (generateClaudeMd, scaffold, skills, agents)
- Analyzer module (6-dimension scoring, autoOptimize, optimizeForSize)
- Headless benchmarking (claude -p integration with proof chain)
- Empirical validation (validateEffect with Pearson r, Spearman rho,
Cohen's d, content-aware executors, multi-trial support)
- Update test count: 1,290 tests across 26 files
- Update module count: 31 modules across 9 layers
- Add generators and analyzer to API quick reference import map
- Add full API reference tables for both modules
https://claude.ai/code/session_01HLi9WmfvFgkDLJxLKGAUTf
2026-02-01 21:54:54 +00:00
e06f273a1b
feat: add CLAUDE.md generators, analyzer, and auto-optimizer
...
Generators (src/generators.ts):
- generateClaudeMd() - structured CLAUDE.md from project profile with
language-specific invariants (TS, JS, Python, Rust, Go, Java) and
framework rules (React, Next.js, Express, Django, etc.)
- generateClaudeLocalMd() - CLAUDE.local.md for personal overrides
with machine notes, editor prefs, and experimental rules
- generateSkillMd() - Claude Code skills with YAML frontmatter
- generateAgentMd() - Agent definitions with YAML frontmatter
- generateAgentIndex() - Agent index.yaml grouped by category
- scaffold() - Full .claude/ directory structure generation
Analyzer (src/analyzer.ts):
- 6-dimension scoring (0-100 each, weighted composite):
Structure (20%), Coverage (20%), Enforceability (25%),
Compilability (15%), Clarity (10%), Completeness (10%)
- benchmark() - before/after comparison with per-dimension deltas
- autoOptimize() - iteratively applies high-priority suggestions
with concrete patches, re-scores to verify improvement
- formatReport() / formatBenchmark() - human-readable output
Analysis of existing CLAUDE.md files:
- Root CLAUDE.md: 73/100 (C) -> 77/100 after optimization
- v3/CLAUDE.md: 64/100 (D) -> 78/100 after optimization
- Key findings: low rule density, oversized constitution section,
missing Architecture section
Tests: 141 new tests (68 generators + 73 analyzer), 1229 total passing
https://claude.ai/code/session_01HLi9WmfvFgkDLJxLKGAUTf
2026-02-01 20:48:36 +00:00
e5ca9411cd
docs: add comprehensive user documentation with guides, tutorials, and diagrams
...
Ground README and guides in how Claude Code actually loads memory:
- CLAUDE.md as team-shared, committed guidance
- CLAUDE.local.md as private, machine-specific overrides
- Upward directory search and monorepo layering
- @import pattern for git worktree compatibility
- /memory command for verification
- Optimizer lifecycle: local experiments promote to root
README additions:
- "How Claude Code Memory Works" section with audience table
- "How CLAUDE.md Becomes Enforceable Policy" mermaid diagram
- "What Goes Where" section with concrete examples
- Updated test count to 1,088 across 24 files
Getting-started additions:
- Background section explaining memory file loading
- Both files in minimal setup example
- Detailed CLAUDE.md vs CLAUDE.local.md guidance
- @import worktree pattern
- Optimizer promotion workflow
- Verification steps with /memory command
Architecture additions:
- "Starting Point" section explaining the gap raw text leaves
- File discovery diagram showing upward search + @import
- Annotated compile/run/evolve lifecycle
https://claude.ai/code/session_01HLi9WmfvFgkDLJxLKGAUTf
2026-02-01 20:23:48 +00:00
176f6ad235
docs: add comprehensive user documentation with guides, tutorials, and diagrams
...
13 documentation files covering all 20+ modules:
Guides:
- Getting Started: installation, setup, core loop
- Architecture Overview: 7-layer design with mermaid diagrams
- Multi-Agent Security: threats, collusion, quorum, trust
- Capabilities and Trust: permission algebra, trust tiers, authority
- Knowledge Management: uncertainty, temporal, truth anchors
- WASM Kernel: building, using, benchmarking
Tutorials:
- Enforcement Gates: wire gates into hooks
- ContinueGate: prevent runaway agent loops
- Proof & Audit Trail: tamper-evident chains
- CI/CD & Headless Testing: automated compliance
Reference:
- API Quick Reference: all exports, methods, types
Diagrams:
- 10 mermaid diagrams: system overview, enforcement pipeline,
decision trees, trust flow, capability delegation, memory
security layers, knowledge stack, proof chain, meta-governance
lifecycle, compile-retrieve-gate sequence
https://claude.ai/code/session_01HLi9WmfvFgkDLJxLKGAUTf
2026-02-01 19:55:55 +00:00
2dd2bb891f
feat: add Rust WASM policy kernel with SIMD128 acceleration
...
Two-layer architecture: Layer A (Rust→WASM kernel for pure functions) +
Layer B (Node host bridge with transparent JS fallback).
Kernel modules:
- proof: SHA-256, HMAC-SHA256, content hashing, chain verification
- gates: secret scanning (8 patterns), destructive command detection (12 patterns)
- scoring: shard relevance scoring with domain/risk/keyword bonuses
- batch: 9-operation batch API to minimize WASM boundary crossings
Performance (SIMD128 enabled, 10k events):
- SHA-256 throughput: 1.80x faster (910k/s vs 505k/s)
- Secret scan (dirty): 1.96x faster (362k/s vs 185k/s)
- Secret scan (clean): 1.68x faster (676k/s vs 402k/s)
- Proof chain: 1.25x faster (61ms vs 76ms)
Includes:
- Pre-built wasm-pkg/ with SIMD+O2 optimization
- 15 acceptance tests (parity + 10k throughput benchmarks)
- ADR-G025 documenting architecture and performance
- README updated with WASM kernel docs and KPIs
- Wired exports into index.ts and package.json
https://claude.ai/code/session_01HLi9WmfvFgkDLJxLKGAUTf
2026-02-01 17:44:35 +00:00
a57ea60aab
feat: add ContinueGate, wire all exports, rewrite README
...
- Add ContinueGate (ADR-G024): step-level gate that evaluates whether
long-running agents should continue, using budget slope detection,
rework ratio tracking, and coherence/uncertainty thresholds
- Wire 5 missing wave 3 module exports into index.ts (trust,
truth-anchors, uncertainty, temporal, authority)
- Add 8 missing sub-export paths to package.json (trust, truth-anchors,
uncertainty, temporal, authority, adversarial, meta-governance,
continue-gate)
- Rewrite README with product-oriented introduction, mermaid architecture
diagram, 28-module reference table, ship phases, acceptance tests,
decision matrix, and failure modes
https://claude.ai/code/session_01HLi9WmfvFgkDLJxLKGAUTf
2026-02-01 17:05:45 +00:00
ff4a4b5fd1
docs: add wave 3 ADRs G017-G023 and update README index
...
- G017: Trust Score Accumulation (tiers, decay, 5:1 penalty/reward)
- G018: Truth Anchor System (immutable HMAC-signed facts, conflict resolution)
- G019: First-Class Uncertainty (confidence intervals, evidence, contested detection)
- G020: Temporal Assertions (bitemporal windows, supersession, reasoning)
- G021: Human Authority and Irreversibility (escalation hierarchy, proof levels)
- G022: Adversarial Model (threat detection, collusion, memory quorum)
- G023: Meta-Governance (constitutional invariants, amendments, optimizer bounds)
README ADR index updated from Proposed to Accepted with file links.
https://claude.ai/code/session_01HLi9WmfvFgkDLJxLKGAUTf
2026-02-01 16:29:38 +00:00
ee3c182b79
feat: complete wave 2 with conformance kit, exports, ADRs, and README
...
- Add conformance-kit.ts: Memory Clerk acceptance test with SimulatedRuntime,
MemoryClerkCell, and ConformanceRunner for replay verification
- Add manifest-validator.test.ts: 59 tests for fails-closed admission
- Add conformance-kit.test.ts: 42 tests for cell runtime and replay
- Wire all wave 2 modules into index.ts (capabilities, conformance-kit,
proof, memory-gate, coherence)
- Add 8 new export paths to package.json (proof, memory-gate, coherence,
artifacts, capabilities, evolution, conformance-kit)
- Write ADR-G010 through G016 covering capability algebra, artifact ledger,
manifest validator, evolution pipeline, conformance kit,
coherence-driven throttling, and agentic container integration
- Add package README with quantified impact, API reference, tutorials,
architecture diagram, and full test matrix
- 639 tests passing across 17 test files
https://claude.ai/code/session_01HLi9WmfvFgkDLJxLKGAUTf
2026-02-01 12:04:55 +00:00
0add23b8d7
docs: add ADR-G001 through G005 for guidance control plane
...
Architecture Decision Records for core design decisions:
- G001: Why a separate control plane beside Claude Code
- G002: Constitution/shard split for token-efficient retrieval
- G003: Weighted pattern matching for intent classification
- G004: Four enforcement gates (destructive, allowlist, diff, secrets)
- G005: Cryptographic proof envelopes for run auditing
Remaining ADRs (G006-G009) in progress.
https://claude.ai/code/session_01HLi9WmfvFgkDLJxLKGAUTf
2026-02-01 11:32:56 +00:00
rUv