Skip to content

Latest commit

 

History

History
138 lines (105 loc) · 6.78 KB

README.md

File metadata and controls

138 lines (105 loc) · 6.78 KB

DocGuard — CDD Enforcement Extension for Spec Kit

Enterprise-grade Canonical-Driven Development (CDD) enforcement and AI-readable project memory for Spec Kit. DocGuard builds source-derived documentation context for supported code (generate --plan), refreshes generated sections as code changes (sync), and checks configured rules (guard) — with deterministic mechanical fixes (fix --write) where it can and grounded agent prompts where prose is needed.

Features

  • Configurable validators — Structure, Security, Doc Quality, Test-Spec, Drift-Comments, API-Surface, Freshness, Cross-Reference, and more; use docguard --help for the current surface
  • Language-specific extraction — Supported checks cover several languages and monorepos. Coverage varies by detector; unsupported inputs remain unverified.
  • AI-powered Generate — generate --plan builds the code-truth skeleton in <!-- docguard:section --> markers and emits a structured agent task manifest; the AI writes the prose.
  • Refresh and review — sync surgically refreshes code-truth doc sections in place, preserves human prose, flags prose for agent review.
  • Mechanical fix --write — deterministic, no-LLM: remove stale documented endpoints, refresh stale "N validators" counts, replace stale version refs, insert missing ## [Unreleased].
  • 5 AI Skills — docguard-fix, docguard-guard, docguard-review, docguard-score, docguard-sync (enterprise-grade behavior protocols, not just step-lists)
  • Workflow Chaining — YAML handoffs enable guard → sync → fix → review → score flows
  • Spec Kit Hooks — Quality gate integrations at implement, tasks, and review phases
  • Minimal Dependencies — one pinned, optional-load parser (@babel/parser); Node.js built-ins otherwise

Installation

npm install -g docguard-cli

Requires Spec Kit ≥ 0.11.2 (requires.speckit_version). docguard init registers this extension from the installed package and re-registers it when its version differs from the CLI's; docguard upgrade --apply does the same.

The command and skill files run docguard from PATH when it is installed, and otherwise npx --yes docguard-cli@<version>, pinned to the release they ship with. They never fetch @latest. The orchestration scripts require a local node_modules/docguard-cli installation or docguard on PATH; they do not download a CLI.

Quick Start

# Initialize CDD in your project
docguard init

# Check documentation health
docguard guard

# Get AI-ready fix prompts
docguard fix --doc architecture

# Calculate maturity score
docguard score

Commands

Agents invoke these in their own form: /speckit-docguard-guard in Claude Code and other skills-based agents, /speckit.docguard.guard for the generic integration. Spec Kit registers no extension command for generic, so docguard init writes them into the generic commands directory. The alias column lists DocGuard's own commands, installed only for generic or an agent DocGuard cannot place.

Command Alias Purpose
speckit.docguard.init — Initialize CDD in a project
speckit.docguard.guard docguard.guard Run configurable quality gate with severity triage
speckit.docguard.fix docguard.fix AI-driven documentation repair with codebase research
speckit.docguard.review docguard.review Cross-document semantic consistency analysis (read-only)
speckit.docguard.score docguard.score CDD maturity score with ROI improvement roadmap
speckit.docguard.diagnose — Diagnose issues + generate multi-perspective AI prompts
speckit.docguard.generate — Reverse-engineer canonical docs from codebase
speckit.docguard.sync — Refresh code-truth sections and flag prose for review
speckit.docguard.trace — Generate requirements traceability matrix
speckit.docguard.brief — Load current spec intent before specification
speckit.docguard.preflight — Gate the generated spec before task generation
speckit.docguard.complete — Plan reviewed completion and regenerate active context after verification

AI Skills

DocGuard provides 5 enterprise-grade AI behavior protocols modeled after Spec Kit's skill architecture:

Skill Lines What It Does
docguard-guard 155 6-step execution with severity triage, structured reporting, remediation recommendations
docguard-fix 195 7-step research workflow with per-document codebase research, 3-iteration validation loops
docguard-review 170 Semantic cross-document analysis with 6 analysis passes and quality scoring matrix
docguard-score 165 CDD maturity assessment with ROI-based improvement roadmap and grade progression
docguard-sync — Refresh code-truth sections while preserving human prose and routing it for review

Skills differ from commands in a critical way: commands tell agents what to run (step-lists), while skills tell agents how to think, validate, and iterate (behavior protocols).

Spec Kit Integration

Workflow Hooks

DocGuard integrates into the spec-kit workflow through hooks:

hooks:
  before_specify:    # Mandatory — read current spec intent first
    command: speckit.docguard.brief
  after_implement:   # Guard is mandatory; completion review is optional
    - command: speckit.docguard.guard
    - command: speckit.docguard.complete
  after_converge:    # Optional completion review after convergence
    command: speckit.docguard.complete
  before_tasks:      # Mandatory — gate the generated spec
    command: speckit.docguard.preflight
  after_tasks:       # Optional — show score after tasks
    command: speckit.docguard.score

docguard init reports the hooks as active only after each mandatory hook resolves to a command file the agent can run.

The hooks call the deterministic CLI contract. The pre-specification hook emits a current-intent briefing; the pre-task hook checks the actual generated spec. Spec Kit dispatches the hooks through its agent workflow, while docguard specs --check remains the CI enforcement surface.

Workflow Chaining

All commands support YAML handoffs for seamless workflow chaining:

guard → fix → review → score
  ↑                      ↓
  └──────────────────────┘

Scripts

Script Purpose
docguard-check-docs.sh Discover docs, return JSON inventory with metadata
docguard-suggest-fix.sh Run guard, prioritize fixes as JSON
docguard-init-doc.sh Initialize canonical doc with metadata header

All scripts support --json mode for AI-parseable output.

License

MIT © Ricardo Accioly