Tools · Harness-wrapped agent skills

Install a skill. Run it inside the harness.

Every skill installs into any of 20+ MCP agents from one paste and runs as the tool skill.<name> inside the HyperMove harness — observability error-capture, sentinel policy, and output-enforcement. Machine-readable at /api/skills.

Generate a tool from any URL

Paste a dApp link → scan → get an agent-callable MCP config.

Open Generator

Harness primitives

The runtime every skill runs inside — adopt one capability at a time.

hm-error-handler

skill.hm-error-handler
free

Structured error capture + trace + recursive self-heal feedback for any wrapped call.

error-handlersentinelt1_read

Composes: lib/observability

SKILL.md

hm-output-enforcer

skill.hm-output-enforcer
free

Enforce-don't-instruct verify-gate: check arbitrary output against required fields; report pass/fail.

error-handlersentinelt1_read

Composes: lib/harness/output-enforcer

SKILL.md

hm-doc-extract

skill.hm-doc-extract
$0.02/page

Parse PDF/image/Office → structured data with an auto-generated or provided extraction schema (Upstage).

error-handlersentineloutput-enforcerdoc-extractt2_realtime

Composes: Upstage Document Parser, lib/harness/output-enforcer

SKILL.md

hm-local-mcp

skill.hm-local-mcp
free / $49-mo managed

Privacy-first local FastMCP server config — sensitive documents never leave the premises.

error-handlersentinelt1_read

Composes: FastMCP 3.0 (local + Skills provider)

SKILL.md

hm-inference-router

skill.hm-inference-router
5% of savings / $19-mo

Route simple queries to a 99%-cheaper open model (DeepSeek V4), hard ones to a frontier model; report est. savings.

error-handlersentinelt1_read

Composes: LiteLLM, DeepSeek V4

SKILL.md

hm-search-harness

skill.hm-search-harness
$0.001/query

Goal-framing search: force a one-sentence goal, run parallel retrieval, cut RAG "caveman-query" tokens ~95%.

error-handlersentinelt1_read

Composes: Mixedbread goal-framing

SKILL.md

Business models

Productized playbooks composed from the primitives.

hm-digital-sdr

skill.hm-digital-sdr
free to try

Digital SDR in a Box — scrape, enrich, draft personalized outreach under a strict $20/day spend cap.

error-handlersentineloutput-enforcert2_realtime

Composes: hm-inference-router, hm-output-enforcer, sentinel:maxSpendPerDay=20

SKILL.md

hm-smart-inference-router

skill.hm-smart-inference-router
free to try

Smart Inference Router (pro) — compress a client's search/support loops; outcome-gated on a live 70%+ token-bill cut.

error-handlersentinelt2_realtime

Composes: hm-inference-router, lib/observability

SKILL.md

hm-oracle-gap-search

skill.hm-oracle-gap-search
free to try

Oracle-Gap Search Harness — kill RAG context-bloat with goal-framing + parallel search (~95% token cut).

error-handlersentineloutput-enforcert3_vector

Composes: hm-search-harness, hm-output-enforcer

SKILL.md

hm-sentinel-precommit

skill.hm-sentinel-precommit
free to try

Sentinel pre-commit hook + Cognitive Debt Sentry — block uncompiled/untested agent commits; feed failures back.

error-handlersentinelt2_realtime

Composes: hm-output-enforcer, sentinel, hm-error-handler

SKILL.md

hm-ap-desk-parser

skill.hm-ap-desk-parser
free to try

Offline AP desk — parse invoice PDFs → validated CSV, math-check line-items sum to total, zero data egress.

error-handlersentineloutput-enforcerdoc-extractt2_realtime

Composes: hm-doc-extract, hm-local-mcp, hm-output-enforcer

SKILL.md

hm-brand-ad-studio

skill.hm-brand-ad-studio
free to try

YC-style Brand & Ad-Creative Studio — WebGL shader renders + Playwright 4-second perfect-loop video ads.

error-handlersentineloutput-enforcert2_realtime

Composes: hm-output-enforcer, sentinel, WebGL, Playwright

SKILL.md

xrpl-search

skill.xrpl-search
free

Search the latest XRPL developer resources — docs, XLS standards, XRPLF repos, Ripple insights — to build products. Free, 10 queries/24h.

error-handlersentinelt1_read

Composes: lib/mcp/search, lib/mcp/xrpl-sources

SKILL.md

xrpl-research-pro

skill.xrpl-research-pro
free to try

Deep, web-wide, freshness-ranked XRPL research — neural search + synthesis + structured findings, powered by Exa. Free to try (metered by the shared free-call tier; deep-reasoning depth is cost-guarded by FEATURE_XRPL_DEEP_REASONING).

error-handlersentineloutput-enforcert3_vector

Composes: Exa, lib/mcp/exa-client, lib/harness/output-enforcer

SKILL.md