v1.6.0: Source Exclusions and Streamlined Wiki Governance

Project Wiki v1.6.0

Project Wiki v1.6.0 introduces deterministic repository source exclusions, compact weekly audit logs, and streamlined instructions with fewer redundant procedures and navigation steps.

What's New

• Added .project-wiki/.wikiignore to control which repository sources participate in wiki discovery, synchronization, and document intake.
• Added scripts/wiki_scope.py for source listing, reading, searching, and diffs with automatic exclusion filtering.
• Replaced monthly audit logs with ISO-week files: logs/wiki-log-YYYY-Www.md.
• Added cumulative audit entries for consecutive updates to the same activity.
• Consolidated shared policies, reconciliation, and logging instructions while preserving existing review and provenance guarantees.

Ingestion & Review Pipeline

• Inbox discovery, source hashing, document ingestion, and raw-source validation now respect .wikiignore.
• Excluded sources are filtered before their content is returned to the agent.
• Existing wiki records, extracted intake artifacts, and audit history remain consultable without refreshing them from excluded sources.
• Exclusions alone do not trigger missing-source or drift findings.
• Document ingestion now has one consolidated operational procedure, with implementation details available in the same reference.
• Complete chunk review, bidirectional requirement evidence, and ledger SHA-256 checkpoints remain unchanged.

Deterministic Governance & Scaffolding

New and improved helpers provide:

• Automatic creation of a comments-only .wikiignore during new-wiki scaffolding.
• Repository-root-relative patterns, directory exclusions, wildcards, and ! inclusion exceptions.
• Consistent source filtering across discovery, reads, searches, diffs, and intake.
• Structural validation of wiki records while skipping existence, content, and hash checks for excluded raw sources.
• Source-processing failures when exclusion rules cannot be evaluated reliably.

Wiki audit logging now uses ISO weeks and compact cumulative entries:

• Extend only the latest entry when the same activity continues within the current week.
• Preserve entry IDs, dates, anchors, and previous information.
• Keep initialization, migration, document integration, and other mandatory audit events separate.
• Capture useful corrections and alert/open-question outcomes without empty fields or duplicated metadata.

The canonical machine-readable contract remains in schema/project-wiki.yml, now at version 1.6.0.

Documentation

Instructions have been streamlined to reduce redundant content and unnecessary navigation:

• Shared finalization, open-question reconciliation, and audit logging are consolidated in common-policies.md.
• Always-on instruction bootstrap lives in initialization-workflows.md.
• Requirement authoring rules live alongside the requirements templates.
• Document ingestion procedures and technical details are maintained together.
• Source inbox processing and repository exclusion configuration have dedicated references.

Migration

This release updates the wiki schema from 1.5.1 to 1.6.0.

After upgrading, run:

/project-wiki maintain

Maintenance adds .project-wiki/.wikiignore when missing, preserves existing exclusion rules and project history, and aligns local templates and instructions with the current schema.

To configure exclusions before the first repository scan, run init, edit .project-wiki/.wikiignore, then run scan. A first scan without configured rules proceeds without exclusions.

v1.5.1 — Canonical Scaffolding, Advanced Review Progress & Atomic Traceability

Project Wiki v1.5.1

Project Wiki v1.5.1 introduces automated canonical scaffolding, structured document inspection and auditing, machine-facing atomic requirement evidence, and a fully modularized workflow architecture.

What's New

• Added scripts/wiki_scaffold.py to generate and validate the fixed canonical wiki skeleton atomically in staging with an exclusive no-replace publish operation.
• Added traceability/requirement-evidence.yml as the machine-facing source of truth for atomic requirement provenance (REQ-*, NFR-*, CON-* ↔ DOCIN-*-CH-*).
• Major upgrade to scripts/review_progress.py with inspect, view, apply, audit, and audit-skips subcommands.
• Integrated intake ledgers and requirement-evidence.yml must now reconcile bidirectionally.
• Requirement markdown bodies no longer contain inline chunk paths or extraction markers, keeping readable specifications clean.
• Ledger audits enforce cryptographic SHA-256 checkpoints (--expect-ledger-sha256) to reject stale or out-of-order mutations.
• Streamlined SKILL.md for compact routing and reduced agent context overhead.

Ingestion & Review Pipeline

• Structured document inspection (inspect) generates hierarchical section outlines (SEC-NNN), word counts, and progress status before review begins.
• Source reading (view) provides clean, full-document (--all), section-bounded (--section), or targeted chunk reread (--chunks) output without wrapper noise.
• Ledger auditing (audit) verifies complete chunk coverage (review-complete) and records exact ledger SHA-256 digests.
• Added audit-skips diagnostic to inspect skipped chunk sequences and scan for potential missed normative language.
• Requirement topic scaling strictly separates functional, non-functional, and constraint records into their designated topic files and sidecars.

Deterministic Governance & Scaffolding

New and improved helpers provide:

• Atomic new-wiki scaffolding with exclusive staging and recipe validation (wiki_scaffold.py)
• Extraction inspection, progressive source viewing, and atomic ledger updates (review_progress.py)
• Strict validation of requirement-evidence.yml format, duplicates, and reciprocal target mappings (validate_wiki.py)
• Repository-relative source path resolution and anchor integrity verification
• Contract drift checks for semantic paths, scaffold recipes, and v1.5.1 schema integrity

The canonical machine-readable contract now lives in schema/project-wiki.yml (version 1.5.1).

Documentation

Workflow documentation has been completely modularized. Monolithic workflow guides have been decomposed into focused, single-purpose references:

• initialization-workflows.md
• update-workflows.md
• document-ingestion.md
• maintenance-workflows.md
• synchronization-workflow.md
• automatic-workflows.md
• common-policies.md
• wiki-structure.md

Migration

This release updates the wiki schema from 1.4.0 to 1.5.1.

After upgrading, run:

/project-wiki maintain

v1.4.0: Verifiable Document Review and Deterministic Governance

Project Wiki v1.4.0

Project Wiki v1.4.0 makes document ingestion, large requirements reviews, and wiki governance deterministic and verifiable.

What's New

• Added review-progress.yml, an exhaustive review ledger containing one entry for every document chunk.
• Added scripts/review_progress.py to select bounded review batches and apply atomic progress updates.
• Intakes cannot be marked reviewed or integrated while chunks remain incomplete.
• Classified chunks in integrated intakes must reference their registered wiki target IDs.
• Skipped chunks require an explicit reason.
• Hints now prioritize review order but never exclude chunks from coverage.
• One chunk may produce multiple requirements, decisions, questions, or target IDs.

Ingestion Improvements

• Document ingestion is now transactional, with staging, atomic publication, rollback, immutable source snapshots, and optional preflight hash enforcement.
• Chunking preserves headings, line breaks, paragraphs, and stable word limits.
• DOCX paragraphs and tables are extracted in their actual document order.
• Large documents are stored completely as progressively readable chunk files without loading the entire source into context.

Deterministic Governance

New and improved helpers provide:

• Source inbox duplicate and history checks
• Transactional duplicate quarantine
• Strict schema loading with duplicate YAML key detection
• Wiki tree, frontmatter, ID, registry, path, link, and provenance validation
• Review ledger coverage and integration validation
• Documentation and template drift detection

The canonical machine-readable contract now lives in schema/project-wiki.yml.

Documentation

Workflow and template documentation was reorganized for progressive disclosure. Large monolithic files are now compact routing indexes pointing to focused references and template catalogs.

Migration

This release updates the wiki schema from 1.3.0 to 1.4.0.

After upgrading, run:

/project-wiki maintain

v1.3.0: Durable Project Memory for AI Coding Agents

📖 Project Wiki v1.3.0

Project Wiki is an IDE-neutral agent skill that brings Andrej Karpathy's LLM Wiki pattern into software repositories by creating and maintaining an agent-readable, durable knowledge base in .project-wiki/.

🔑 Key Features

• Progressive Disclosure: Agents start at INDEX.md and load only the minimal files needed, protecting the context window from token bloat.
• Zero-Friction Agent Lifecycle: Automatic context preflight before coding and automatic documentation/traceability updates after code edits.
• Bidirectional Traceability: Permanent IDs (REQ-*, ADR-*, CR-*) linked directly to repository source paths.
• Document Intake: Ingestion helper for PDF, DOCX, and Markdown requirements into structured intake/ artifacts.
• Active Governance: Continuous semantic linting and persistent risk alerts.

🌐 Cross-Platform Compatibility

Works seamlessly with:

• GitHub Copilot (VS Code / JetBrains)
• Claude Code
• Antigravity
• Codex & OpenCode
• Cursor

📦 What's New in this Release

• Added complete community & contributor governance (CONTRIBUTING.md, CODE_OF_CONDUCT.md, SECURITY.md).
• Added interactive GitHub Issue forms (bug_report.yml, feature_request.yml) and Pull Request template.
• Refined documentation.