- Home
- Skills
- Agents & Orchestration
- AgentHarness Toolkit: All 11 Agent Skills in One
Works with the AI Agents you already use
AgentHarness Toolkit: All 11 Agent Skills in One
**What you get.** One skill with a short router and 11 complete guides.
$49
AgentHarness Toolkit: All 11 Agent Skills in One
Example session with this skill installed
I'm about to add a payment SDK and a migration for a new payments table, then open a PR. Help me do it properly.
- Read your context and instructions
- Compiled the agentharness toolkit
Example session (illustrative).
Guides selected (in order): dependency-vetting → zero-downtime-migrations → agent-harness-gate → pr-ready. I'll say which one I'm following at each step.
1. dependency-vetting
stripeexists on npm, first published in 2011, repo link matches github.com/stripe/stripe-node, MIT license.npm view stripe dependencies: none, so +0 transitive packages;npm install --ignore-scriptsin a scratch dir showed no lifecycle scripts.npm audit --omit=dev: 0 vulnerabilities.- Decision: add, pinned in
package-lock.json. PR note drafted.
2. zero-downtime-migrations
- New table only, so no lock risk on existing data. The foreign key to
orders(90M rows) is added asNOT VALIDin its own migration, validated in a second one, each withlock_timeout = '3s'. - The index on
payments.order_idis created in the same migration as the table, since the table is empty. - Rollback plan written; nothing run against any shared database.
3. agent-harness-gate
harness run --agent claude-code --intent "add payments table and stripe sdk" --json: exit
1 (blocked). Blocking reason: new dependency stripe does not match the repo's [security.dependencies] allow list in harness.toml.
- I did not edit the allowlist myself. You approved adding it; re-run: exit 0,
"verdict": "Pass".
4. pr-ready
make ci: exit 0, 312 passed. Self-review removed oneconsole.log.- 480 changed lines in one logical change, so one PR. Title:
feat(payments): add payments table and stripe client. Description lists the exact commands above and their results, the migration rollback steps, and the new dependency note.
Waiting for your OK to push the branch and open the PR as a draft.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem. Coding agents are fast and confident, and the mistakes cluster in the same places: they say "done" without running anything, merge branches that break each other, install packages nobody vetted, paper over flaky tests, write migrations that lock production, and open PRs nobody can review. Each of those has a careful procedure. Installing eleven separate skills to cover them is a chore.
What you get. One skill with a short router and 11 complete guides. The router reads the situation, picks the one guide that fits, and loads only that guide, so your agent's context stays small.
- Verify before "done": run the AgentHarness gate and fix what it blocks.
- Parallel agents: worktrees,
agent/*branches, semantic conflict checks and a merge order. - CI gate: the AgentHarness GitHub Action or binary, per-stack config and a required check.
- Leaked secret: triage, rotate first, purge history only with your approval, prevent.
- Dependency vetting: typosquats, licenses, install scripts and vulnerabilities, before installing.
- Flaky tests: reproduce with numbers, classify the cause, fix, quarantine with a paper trail.
- Offline tests: turn network-calling tests into hermetic ones without losing coverage.
- Zero-downtime migrations: PostgreSQL and MySQL changes that don't lock live tables.
- Regression bisect: find the commit that broke it with a correct
git bisect runscript. - Agent rules: AGENTS.md and CLAUDE.md built from commands that were actually run.
- PR ready: real checks, stacked PRs without interactive git, honest descriptions.
When several apply, the router runs them in a sensible order: emergencies first, vetting before installing, verification and PR packaging last.
Every command was tested. Each guide's commands were run in real scratch projects before release, and anything that could not be run here (Go, MySQL, cloud CLIs) was checked against official docs.
Who it's for. Anyone letting Claude Code, Cursor, Codex CLI, Copilot or similar agents work in real repositories who wants the whole set at once.
Requirements. git and your project's own tooling. The gate, branch-coordinator and CI-setup guides need the free, open-source harness CLI from github.com/KadonWills/agent-harness (a few other guides mention it as an optional extra check); the crates.io package named harness-cli is unrelated. The other guides work without it.
Requires the harness CLI from GitHub for gating guides. Does not support non-relational databases. Limited to npm, PyPI, crates.io, and Go for dependency vetting.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 17 days ago
Requires
Permissions
Allowed Hosts