5dive skills

Skills published by 5dive for use with skills.sh and any agent harness that loads SKILL.md bundles (Claude Code, openclaw, hermes, …).

Skill	Purpose
5dive-cli	Drive the local 5dive CLI on a 5dive runtime VM. Lets an agent spawn, inspect, send to, and tear down sibling agents on the same host.
diagnose	Self-diagnosis for 5dive agents. Runs auth, service, disk, memory, and skill-integrity checks on the VM and returns a root-cause report — so the agent can fix problems itself instead of asking the user.
compile-knowledge	Compile durable knowledge into interlinked-markdown stores the "karpathy method" way — atomic files, [[links]], a maintained index — so the agent gets smarter over time instead of relearning. Memory-first; shared wiki only for teams.
openagent	Author your own OpenAgent persona and mint your shareable holo trading card. Self-service: write a <id>.persona.yaml, validate it, check your rarity tier (Common→Mythical) + completeness, render the PNG card, and PR into the character-packs registry.
copywriting	Write, rewrite, and improve conversion-focused marketing copy for any page — homepage, landing, pricing, feature, product. Headlines, CTAs, value props, taglines, hero sections.
no-ai-slop	Edit drafts into sharper, more human writing while preserving the writer's voice, or detect AI-slop patterns without rewriting. Use to make a draft clearer, more direct, and less AI-sounding, or to check whether writing reads as AI. MIT, courtesy of Peter Yang.
ad-creative	Generate and iterate ad creative at scale — headlines, descriptions, and primary text for paid platforms — and refine based on real performance data.
find-loops	Search the agenticloops.dev directory for an existing agentic loop (a recurring, scheduled agent) and install it. The loop-level analogue of find-skills.
loops	Author or modify a portable LOOP.md — a trigger + skills + a prompt in one file any harness can install and run on a schedule. The loop-level analogue of skill-creator; to install a loop that already exists use find-loops.
playwright-e2e	End-to-end testing and click-verification of web apps with Playwright. Install, write specs, drive authenticated pages, screenshot, and run in CI. Use to prove a web change actually works in a real browser before shipping.
code-review	Review a diff or pull request for correctness first, quality second. Logic bugs, edge cases, error handling, races, missing tests, API misuse, performance, readability, with a severity model and clear approve vs request-changes verdict.
verify	Grade a delivered claim against the artifact instead of against the report of it. Split a claim into checkable assertions, grade each against a named instrument, and emit a three-state verdict — pass, fail, or not-reached — with what you did not check stated explicitly.
deep-research	Run structured, multi-source research that produces a trustworthy, cited synthesized answer rather than a pile of links. Scope the question, fan out across sources, judge source quality, adversarially verify each key claim, separate fact from inference, and report with explicit confidence.

Quick start (any coding agent)

Install straight from the skills.sh directory with the open skills CLI — works in Claude Code, Codex, Cursor, and any harness that loads SKILL.md:

# mint your own OpenAgent persona + shareable holo card
npx skills add 5dive-ai/skills --skill openagent

# find, install, and run agentic loops (recurring agents)
npx skills add 5dive-ai/skills --skill loops

# or pull the whole set
npx skills add 5dive-ai/skills --skill '*'

openagent and loops are the best starting points — they work in any agent, no 5dive account required. loops pairs with the open agenticloops.dev directory + LOOP.md spec.

Install on a 5dive agent

From the dashboard: Agents → Connect skills → 5dive-cli → Install.

From the host:

sudo 5dive agent skill <agent-name> add \
  --source=5dive-ai/skills \
  --skill=5dive-cli

Or via the upstream skills CLI from any agent's home:

npx skills add https://github.com/5dive-ai/skills --skill 5dive-cli --agent claude-code --yes

License

MIT.

Description budget

A skill's name + description is loaded at selection time, in every session, whether or not the skill fires. A command catalog there is paid on every turn and mis-routes the selector, so every description in this repo is capped at two discriminating sentences — the capability, the activation condition, and at most one exclusion. Command names, examples and feature catalogs live in the skill body or in references/.

./scripts/check-descriptions.sh          # enforces the cap, prints total chars

evals/routing.json holds the behavioural routing cases that guard the boundaries the cap is there to protect — an unrelated failed command must not load diagnose, a normal local coding task must not load 5dive-cli, searching for a scheduled workflow selects find-loops while authoring a LOOP.md selects loops.