Agent Catalog

The full list of subagents that ship in agents/. Tier × tools × spawn-source for each.

Planning

np-planner — tier: opus

Tools: Read, Write, Bash, Glob, GrepSpawned by: np:plan-phase (standard / --gaps / revision mode)

Creates executable phase plans with task breakdown, dependency analysis, and goal-backward verification. Honors CONTEXT.md locked decisions verbatim — never simplifies a D-XX to a "stub". When the phase is too large to honor every decision in one plan, returns a PHASE SPLIT RECOMMENDED to the orchestrator instead of silently scope-reducing.

np-plan-checker — tier: opus

Tools: Read, Grep, GlobSpawned by: np:plan-phase verification loop (D-15)

Adversarial reader of PLAN.md. Walks the plan goal-backward against ROADMAP.md success_criteria and emits a single YAML verdict: status: passed | issues_found plus findings[] keyed to the canonical 10-code taxonomy from docs/agent-frontmatter-schema.md. Does not propose fixes; the orchestrator re-spawns the planner in revision mode.

np-researcher — tier: sonnet

Tools: Read, Write, Bash, Grep, Glob, WebSearch, WebFetch, mcp__context7__*, mcp__firecrawl__*, mcp__exa__*Spawned by: np:research-phase (standalone), np:plan-phase (integrated)

Phase-level technical researcher. Probes WebFetch + Context7 availability on startup; if both are absent, enters Offline-Confirm Protocol and produces a local-only RESEARCH.md with a mandatory ## Research Coverage annotation listing what is local-only.

Execution

np-executor — tier: sonnet

Tools: Read, Write, Edit, Bash, Grep, GlobSpawned by: np:execute-phase (per task)

Atomic-commit-per-task executor. One task per spawn, one commit per task (ADR-0004). Reads files_modified verbatim — no auto-discovery, no scope expansion. Routes commits exclusively through np-tools.cjs commit-task so the gitignore-guard runs.

np-verifier — tier: sonnet

Tools: Read, Bash, Grep, GlobSpawned by: np:verify-work

Post-execution goal-backward verifier — the twin of plan-checker, different timing. Reads ROADMAP.md SCs + PLAN.md + task commits + the union of files_modified. Classifies each SC as Pass / Fail / Defer / Needs-User-Confirm. Does not propose fixes, does not edit source.

Review

np-code-reviewer — tier: opus

Tools: Read, Grep, GlobSpawned by: np:code-review

Source-file reviewer. Writes a REVIEW.md sidecar with critical / warning / info findings scored against CLAUDE.md conventions, the ADRs, PROJECT.md constraints, and common security/perf anti-patterns. Supports --depth quick | standard | deep.

np-code-fixer — tier: sonnet

Tools: Read, Write, Edit, Bash, Grep, GlobSpawned by: np:code-review-fix

Auto-fixer for REVIEW.md findings. One atomic commit per successful fix (D-21 exception to one-commit-per-task). Writes REVIEW-FIX.md with all_fixed | partial | none_fixed. --auto iteration cap is 3.

np-security-auditor — tier: opus

Tools: Standard read+grep Spawned by: np:secure-phase

Threat-mitigation auditor. Reads PLAN.md's <threat_model> + the implementation, scores each threat as MITIGATED / PARTIAL / UNMITIGATED, writes SECURITY.md from templates/SECURITY.md.

np-nyquist-auditor — tier: haiku

Tools: Standard read+grep Spawned by: np:validate-phase

For each requirement in phase scope, verifies at least one test observes the implementation directly. Scores COVERED / UNDER_SAMPLED / UNCOVERED. Output via templates/VALIDATION.md.

np-ui-auditor — tier: haiku

Tools: Standard read+grep Spawned by: np:ui-review

Retroactive 6-pillar visual audit (Hierarchy / Typography / Color / Spacing / Density / Interaction). Output: UI-REVIEW.md.

np-eval-auditor — tier: haiku

Tools: Standard read+grep Spawned by: np:eval-review

Retroactive audit of an implemented AI phase. Checks against the AI-SPEC.md evaluation plan. Scores each dimension COVERED / PARTIAL / MISSING. Output: EVAL-REVIEW.md.

UI

np-ui-researcher — tier: sonnet

Tools: Standard read+grep Spawned by: np:ui-phase

Produces UI-SPEC.md design contract. Reads upstream artifacts, detects design-system state, asks only unanswered questions.

np-ui-checker — tier: haiku

Tools: Standard read+grep Spawned by: np:ui-phase

Validates UI-SPEC.md against six quality dimensions. Returns BLOCK / FLAG / PASS verdict — drives the revision loop for ui-phase.

AI integration

np-domain-researcher — tier: sonnet

Tools: Standard read+grep + web/mcp Spawned by: np:ai-integration-phase

Researches the business domain, surfaces domain-expert evaluation criteria, industry-specific failure modes, regulatory context. Feeds the eval-planner.

np-framework-selector — tier: opus

Tools: Standard read+grep + web/mcp Spawned by: np:ai-integration-phase

Interactive framework scoring matrix for AI/LLM stack selection. Produces a ranked recommendation with rationale.

np-ai-researcher — tier: sonnet

Tools: Standard read+grep + web/mcp Spawned by: np:ai-integration-phase

Researches the chosen AI framework's official docs. Writes the Framework Quick Reference and Implementation Guidance sections of AI-SPEC.md.

np-eval-planner — tier: opus

Tools: Standard read+grep Spawned by: np:ai-integration-phase

Designs the structured evaluation strategy: failure modes, eval dimensions with rubrics, tooling, reference dataset. Writes the Evaluation Strategy, Guardrails, and Production Monitoring sections of AI-SPEC.md.