AID — AI-Integrated Development: A complete methodology for human-AI collaborative software development. 12 phases, 5 stages, 11 feedback loops.
Agent Claude Code
Transforms requirements, SPEC, and KB into design output — SPEC sections, typed dependency-ordered task breakdowns, feature decomposition, delivery sequencing, and DESIGN-typed task execution including UX and flow advice.
Agent Claude Code
INTERNAL UTILITY (sub-agent only — do NOT invoke from a skill). Performs one mechanical, schema-bounded operation per dispatch — file extraction, template placeholder-fill, or glob enumeration — returning a markdown table or file with path and line evidence.
Agent Claude Code
The only agent that modifies production code. Implements TASK files following specs and KB conventions across all implementation task types (IMPLEMENT, TEST, REFACTOR, CONFIGURE, MIGRATE, FIX), with mandatory build verification and formal IMPEDIMENT.md escalation.
Agent Claude Code
Conducts adaptive one-question-at-a-time dialogue with human stakeholders to gather requirements, clarify ambiguity, and produce REQUIREMENTS.md or Q&A entries.
Agent Claude Code
Runs final release verification, packages artifacts, creates PRs and release notes, manages releases, and updates the KB on ship. Safety-first, verification-focused.
Agent Claude Code
Routes pipeline findings to the next phase or skill, enforces human gates, dispatches agents with context, and manages parallel execution.
Agent Claude Code
Reads and analyzes code, docs, logs, APIs, and external web sources to produce structured Knowledge Base documents and analysis reports — covering existing-state cataloguing, dependency/integration/convention mapping, telemetry interpretation, security analysis, performance profiling, and web-sourced prior art. Web…
Agent Claude Code
Adversarial quality evaluator. Reviews any artifact (code, tasks, specs, plans, KB docs) against its acceptance criteria, rubric, and KB conventions. Produces the 7-column issue ledger with source and severity tags. Does NOT fix anything; does NOT compute the grade.
Agent Claude Code
Authors user-facing documentation — API docs, changelogs, READMEs, release notes, user guides — and reviews existing docs for quality and accuracy.