drift-watch

Audit and set up a codebase for agentic AI development using the 15-principle manifesto

Full-cycle orchestrator chaining all five phases with gates and controls

Audit and set up a codebase for agentic AI development using the 15-principle manifesto

Full-cycle orchestrator chaining all five phases with gates and controls

Reviews code for architectural compliance and design integrity

Deep semantic analysis of codebase against rules, patterns, and lessons

Injects mistake-derived test requirements into the TDD pipeline

Clusters similar lessons and synthesizes testable patterns

Analyzes completed work to identify what was done and learned

Checks implementation for drift from established constraints

Implements minimal code to pass failing tests

Extracts actionable lessons from work context

Searches and retrieves relevant memory items for context

Matches lessons against existing memory to avoid duplicates

Reviews code for performance issues and resource usage

Route and endpoint audit for authentication, authorization, IDOR, JWT, and CORS vulnerabilities

Audit for PII in logs, verbose error responses, sensitive data in URLs, and overly broad API responses

Dependency audit for vulnerable packages, lockfile changes, postinstall scripts, and supply chain risks

Deep trace analysis for SQL, command, XSS, SSRF, and SSTI injection vulnerabilities

Mandatory core-4 reviewer with P0-P3 severity classification and specialist escalation

Credential and secrets scanning using pattern matching, entropy analysis, and git history checks

Reviews code for unnecessary complexity and over-engineering

Writes final memory items in correct schema format

Reviews test quality, assertions, and edge case coverage

Writes failing tests before implementation exists

Reflect on the cycle and capture high-quality lessons for future sessions

Decompose work into small testable tasks with clear dependencies

Deep research producing structured survey documents for informed decision-making

Multi-agent review with parallel specialized reviewers and severity classification

Develop precise specifications through Socratic dialogue, EARS notation, and Mermaid diagrams

Multi-phase test suite optimization with adversarial review

Team-based TDD execution with adaptive complexity and agent delegation

Reviews code for architectural compliance and design integrity

Deep semantic analysis of codebase against rules, patterns, and lessons

Injects mistake-derived test requirements into the TDD pipeline

Clusters similar lessons and synthesizes testable patterns

Analyzes completed work to identify what was done and learned

Checks implementation for drift from established constraints

Implements minimal code to pass failing tests

Extracts actionable lessons from work context