Skills Reference: Agent Engineering, API Design, Architecture, and Evaluation

Overview

This section documents the skill packs that teach matrix-core agents how to reason about agent engineering, API design, architecture, autonomous operation, and evaluation. The source lives under skills/ in the matrix-core repository and includes 173 ported skills (131 keep + 28 adapt) plus 13 native skills authored for Matrix (6 Paxeer network skills, 7 stack-selection playbooks).

Each skill is a paired asset: a prose SKILL.md defining the human-readable policy and workflow, and an optional SKILL.mtx MatrixScript manifest providing the compiler-facing contract. The manifest declares a required slot.target (ArtifactRef), optional slot.constraints, a typed result, and explicit handling for unknown or ambiguous targets. All compiler-facing manifests use cortex.find to resolve targets and trigger clarification prompts when confidence is too low.

Skill Catalog Structure

The skill catalog is indexed by two files:

Skills carry one of three status labels:

The PORT_MANIFEST.json tracks the full drop, defer, and adapt lists from the porting process.

Shared Manifest Format

All visible SKILL.mtx manifests follow a consistent structure:

• SKILL declares identity and routing fields: id, version, display, description, mcl.verbs, determinism, and seed_policy.
• INPUTS defines the target artifact and optional constraints.
• CORTEX declares the memory categories the skill may read.
• PROCEDURE resolves slot.target with cortex.find, then clarifies if confidence is too low or the target cannot be identified.
• OUTPUTS includes a required slot result: ArtifactRef and optional slot unknowns: Unknown[].
• FAILURE_MODES uses target_not_found, ambiguous_after_clarify, policy_violation, and budget_exceeded.

Across the manifests, the compiler-facing behavior is reproducible: packs are seedable, use per_intent seeding, and are organized around a single target artifact rather than open-ended conversation.

Agent Engineering Skills

Agent Harness Construction

skills/agent-harness-construction/SKILL.md covers action-space design, observation formatting, error recovery contracts, context budgeting, and the hybrid ReAct plus function-calling pattern. Observation fields include status, summary, next_actions, and artifacts. Benchmark metrics include completion rate, retries, pass@1, and pass@3.

The SKILL.mtx manifest supports modify and build verbs with standard target resolution.

Agent Introspection Debugging

skills/agent-introspection-debugging/SKILL.md provides a four-phase self-debugging loop: capture, diagnosis, contained recovery, and report generation. Failure patterns include ECONNREFUSED, 429, stale diffs after writes, and tests that fail after a supposed fix.

The manifest uses the analyze verb and clarifies the target when the request is underspecified.

Agentic Engineering

skills/agentic-engineering/SKILL.md defines the operating model for AI-led implementation work: completion criteria before execution, agent-sized task decomposition, model routing by task complexity, and eval-driven success measurement. Cost tracking covers model, token estimate, retries, wall-clock time, and success or failure.

Agent Payment (x402)

skills/agent-payment-x402/SKILL.md adds autonomous payment execution over x402 with per-task budgets, non-custodial wallets, and spending policy enforcement. The visible example centers on set_policy, check_spending, preToolCheck, and a fail-closed entrypoint. Pre-tool logic handles invalid input, transport failures, tool errors, parse failures, and budget exhaustion.

API Design and Documentation Skills

API Design

skills/api-design/SKILL.md is a REST design guide for resource naming, status codes, pagination, filtering, sorting, error responses, versioning, and rate limiting. It includes interface examples for ApiResponse<T> (with data, meta: PaginationMeta, and links: PaginationLinks) and ApiError (with error, code, message, and details: FieldError[]).

API Documentation

skills/api-documenter/SKILL.md covers OpenAPI 3.1, complete endpoint coverage, request and response examples, error documentation, authentication guides, versioning, and interactive documentation portals. It spans REST, GraphQL, WebSocket, gRPC, webhooks, SDK references, CLI docs, and integration guides.

API Connector Builder

skills/api-connector-builder/SKILL.md is for adding a repo-native connector without inventing a new integration architecture. The workflow: inspect at least two existing connectors, map the house style, narrow the target integration, build in the repo's existing layers, and validate against the source pattern.

Architecture and ADR Skills

Architecture

skills/architecture/SKILL.md documents monorepo structure, platform-specific file naming, workspace import boundaries, and a strict hierarchy that prevents circular dependencies. The analysis protocol: assess scope impact, verify patterns, check architecture integrity, and evaluate performance impact before modifying code.

Architecture Decision Records

skills/architecture-decision-records/SKILL.md captures why decisions were made. It activates when a decision is being chosen or documented, or when someone asks why a past choice was made. The ADR format records the original date for past decisions and keeps superseded decisions linked to their replacements.

Autonomous Operation and Audit Skills

Autonomous Agent Harness

skills/autonomous-agent-harness/SKILL.md describes persistent autonomous agent systems using native crons, dispatch, MCP tools, memory, and task queuing. The architecture shows command and hook surfaces feeding an MCP server layer with services such as memory, github, exa, supabase, and browser-use.

Automation Audit Operations

skills/automation-audit-ops/SKILL.md is an evidence-first operator skill for inventorying automations before changing them. It separates configured, authenticated, recently verified, stale or broken, and missing surfaces. The workflow begins with a real-surface inventory and only then moves to keep, merge, cut, or fix-next recommendations.

Evaluation and Analysis Skills

Agent Evaluation

skills/agent-eval/SKILL.md compares coding agents head-to-head using YAML task definitions and metrics for pass rate, cost, time, and consistency across repeated runs. The workflow is evaluation-first: define tasks, run them, compare metrics.

AI Regression Testing

skills/ai-regression-testing/SKILL.md defines a bug-check workflow with mandatory tests (npm run test and npm run build), sandbox-mode request helpers, and regression proposals for each fixed defect.

AST-Grep Code Analysis

skills/ast-grep-code-analysis/SKILL.md provides structural analysis for complex codebases using AST-based matching for security vulnerabilities, performance issues, and structural patterns.

Skill Index Builder

The skill catalog is generated by tools/skills/build_index.py, which reads SKILL.md frontmatter from every skill directory, cross-references PORT_MANIFEST.json for keep/adapt status, and writes both INDEX.md and INDEX.json. Re-run after editing or adding skills:

python3 tools/skills/build_index.py

Source-Backed Inventory