<essential_principles> <parallel_architecture> Expert research runs in parallel subagents — one subagent per confirmed role. Never run expert reviews sequentially. Parallel execution is the core performance guarantee of this skill. Do not collapse experts into a single pass. </parallel_architecture>

<depth_bar> Every finding must be grounded in evidence — a direct quote, a file:line reference, or a concrete observation from the artifact. Surface-level observations without evidence are not acceptable. The depth bar is: would a senior practitioner in this domain consider this finding actionable and specific? </depth_bar>

<confirmation_gate> The user MUST confirm the expert role list before any research begins. This is not optional. Present the proposal, collect confirmation via ask_user_questions, and proceed only on "Confirm all". Do not skip the gate even if the roles seem obvious. </confirmation_gate>

<no_web_search_in_intake> Role identification is pure in-context LLM inference from the artifact content. No web searches during intake. fetch_page is permitted only to read a submitted URL (the artifact itself) — never for supplementary research or role selection. </no_web_search_in_intake>

<skill_audit_specialization> When the artifact is a skill directory (contains SKILL.md), route to workflows/audit-skill.md — NOT to workflows/intake.md. The skill audit pipeline has its own role taxonomy (references/skill-audit-roles.md), structural pre-validation, maturity scorecard, and phased remediation plan that general intake cannot provide. Role selection for skill audits uses the Role Selection Heuristics table in references/skill-audit-roles.md, not general domain inference. </skill_audit_specialization>

<synthesis_is_not_concatenation> The synthesis phase MUST perform a full pre-synthesis analysis pass (convergence scan, contradiction scan, severity aggregation, maturity scoring, open questions merge) before filling any template section. Filling template sections while simultaneously scanning produces incoherent output (L007). Complete the analysis pass first; hold all results in working state; then do a single-pass template fill. </synthesis_is_not_concatenation> </essential_principles>

For general artifacts (code, documents, architecture, proposals, URLs): Provide a file path, directory path, URL, or paste content directly.

For skill audits (SKILL.md or skill directory): Provide the skill directory path (e.g. ./my-skill/ or ~/.claude/skills/my-skill/) or the skill name (e.g. expert-opinion). The skill audit pipeline runs structural pre-validation, a specialized expert panel drawn from prompt engineering and agentic workflow domains, produces a maturity scorecard (1–5 across 5 dimensions), and delivers a phased remediation plan.

<routing_decision> When the user provides input, determine the route BEFORE loading any workflow or required-reading files. Apply these steps in order:

Step 0 — SKILL.md pre-check (for potential skill-audit routes only): If the input looks like a directory path (ends with /, is a bare word, or user says "audit this skill"): Run: bash find {skill_path} -maxdepth 1 -name "SKILL.md" | head -1 (where {skill_path} is the path provided by the user — substitute before executing) - If SKILL.md found → continue to Step 1. - If SKILL.md NOT found → route to intake (Step 2) immediately. Do NOT load any audit-skill.md required-reading files.

Step 1 — Skill audit route: If, after Step 0, the input: - Refers to a directory containing SKILL.md, or - Is a bare word matching a known skill name (resolved via ~/.claude/skills/{name}/, ~/.agents/skills/{name}/, etc.), or - User says "audit this skill", "review this skill", "check this skill" → Route: skill_audit → workflows/audit-skill.md

Step 2 — General artifact route (everything else): → Route: intake → workflows/intake.md

Ambiguity: If a path could be a skill directory or a general codebase: Ask: "Is this a skill directory or a general codebase?" Then route based on the answer. </routing_decision>