🇧🇷 Versão em português

ai-harness

Write an agent, a skill, a command, a hook or a rule once.
One command installs it into every AI tool you use.

CI PyPI License: MIT Agent Skills AGENTS.md

OpenCode Claude Code Codex Hermes

What it is

Write an agent, a skill, a command, a hook or a rule once and use it in every tool you work with. The portable part follows open standards, Agent Skills and AGENTS.md, while the wiring each tool needs stays in adapters/, so nothing here is tied to a single vendor.

Install

harness asks which tools to install into, and which artifacts to bring, then puts them where each tool looks.

pip install ai-harness-cli
harness

Once installed:

harness update   # bring the latest artifacts and reconcile
harness check    # audit what is installed
harness remove   # take back only what harness installed

Contents

The map of the repository, for writing an artifact or finding where a piece lives.

Path What lives there
agents/ Agents, in the format each tool reads
skills/ Skills in the open Agent Skills layout
commands/ Prompts that each tool exposes as a command
hooks/ Guard scripts the adapters share
rules/ Global rules, linked as AGENTS.md
adapters/ What each tool needs to wire the rest up
docs/ How to author an artifact and how the pieces fit together
harness, ai_harness/ The installer
targets.json Where every artifact lands, per tool
CONTRIBUTING.md How to propose a change and how releases are cut

Catalog

What you get once it is installed.

Skills

Name What it does
dream Reads the day's sessions and suggests what is worth remembering, like the fact that you prefer Postgres.
news-digest Reads your feeds and delivers the day's news as one short message, skipping what you already saw.
task-delegation Hands a heavy task to OpenCode so it runs on the strongest model you have available.
topic-research Researches a topic across Hacker News, Reddit and GitHub, with a link behind every claim.

Agents

Name What it does
code-reviewer Reads your diff and points out bugs, security risks and missing tests. Changes nothing.
news-digest-validator Checks the digest before it reaches you and cuts repeated, obvious or badly translated stories.
proposal-validator Sends your plan to a different model, to catch what the first one talked itself into.

Commands

Name What it does
dream Runs dream and applies the suggestions you approve, one by one.
news-digest Asks for today's digest right now.
review-changes Reviews what you changed, before you open the pull request.
topic-research Researches a topic and writes a report you can check, since every claim cites its source.
validate-proposal Sends the plan on the table for an independent second opinion.

Hooks

Name What it does
protect-secrets Stops the agent from opening .env, key and certificate files, even when you ask by accident.

Rules

Name What it does
global Tells the agent how to work everywhere: ask when unsure, show evidence before claiming it is done.

Integration

This is where each artifact lands.

Artifact Claude Code OpenCode Codex Hermes
skills/<name>/ ~/.claude/skills ~/.config/opencode/skills ~/.agents/skills ~/.hermes/skills
agents/<name>.md ~/.claude/agents ~/.config/opencode/agents Not supported Not supported
commands/<name>.md ~/.claude/commands ~/.config/opencode/commands ~/.codex/prompts Not supported
rules/global.md ~/.claude/CLAUDE.md ~/.config/opencode/AGENTS.md ~/.codex/AGENTS.md Not supported

Tool support

Tool Skills Agents Commands Rules Hooks
OpenCode Yes Yes Yes Yes Adapter
Claude Code Yes Yes Yes Yes Yes
Codex Yes Partial Partial Yes Yes
Hermes Yes No No Workspace No

Partial means Codex has no subagent format, and its prompts document no frontmatter, so only the body carries over. Workspace means Hermes reads rules from the project, so there is no global file to install.

Design rules

  • One job per artifact. name and description stay portable, and any restriction is stated in the body too.
  • Acceptance criteria before the implementation, evidence before calling it done.
  • No credentials, no pinned models, no machine specific paths.

The reasoning behind each one is in docs/authoring.md.

Any other tool that reads Agent Skills and AGENTS.md picks up the skills and the rules on its own. Point it at skills/ and at rules/global.md. The rest depends on what that tool supports.

Made with ❤️ in Curitiba 🌳 ☔️

Project links

Data verified by PyPI on
Data provided by the project maintainers, verified at the time the release was uploaded to PyPI.

Key dates

PyPI data
Data sourced directly from PyPI's database.
  • Released:
Latest release

1 maintainer

PyPI data
Data sourced directly from PyPI's database.
Avatar for brunotfurtado from gravatar.com brunotfurtado

Credits

Author: Bruno Tortato Furtado

GitHub Statistics

Data verified by PyPI on 
The GitHub source repository was provided by the project maintainers and verified by PyPI at the time of upload. Stars, forks, and open issues/PRs are derived from that repository and have not been independently verified.

License expression

MIT
View SPDX License List

Requires

Python >=3.9

Tags

agent-skills agents claude-code codex opencode

Classifiers

Report project as malware

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

ai_harness_cli-0.3.0.tar.gz (91.6 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

ai_harness_cli-0.3.0-py3-none-any.whl (108.9 kB view details)

Uploaded Python 3

File details

Details for the file ai_harness_cli-0.3.0.tar.gz.

File metadata

  • Download URL: ai_harness_cli-0.3.0.tar.gz
  • Upload date:
  • Size: 91.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for ai_harness_cli-0.3.0.tar.gz
Algorithm Hash digest
SHA256 3a2b7f1ea4f685e6f815a7d38786ca554101396791da3fb83e8a636afa4b617c
MD5 5db77b51b76c5d8edebea31bc464820c
BLAKE2b-256 91c14bf6313a33899fc1d3682a9e89c03eb4716d4390a45a21926b8e51a7fc28

See more details on using hashes here.

Provenance

The following attestation bundles were made for ai_harness_cli-0.3.0.tar.gz:

Publisher: release.yml on Bruno-Furtado/ai-harness

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file ai_harness_cli-0.3.0-py3-none-any.whl.

File metadata

  • Download URL: ai_harness_cli-0.3.0-py3-none-any.whl
  • Upload date:
  • Size: 108.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for ai_harness_cli-0.3.0-py3-none-any.whl
Algorithm Hash digest
SHA256 e01ec2a659858189ec8e9d60961e96c7d7a79bf80f234d1f2c58955c0debdfd1
MD5 0d9c30c121ff5e7bfd89612f99182214
BLAKE2b-256 247ad7db781db31e25d24e9fe78c08260242f6131756c099ed12a0a5b7f8b40f

See more details on using hashes here.

Provenance

The following attestation bundles were made for ai_harness_cli-0.3.0-py3-none-any.whl:

Publisher: release.yml on Bruno-Furtado/ai-harness

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page