Skip to main content

agent-skills-bundle

Packaged agent skills for Claude Code and Codex. Skills are SKILL.md folders — a cross-agent standard, so the same skill works in both tools unmodified. install copies them into the directories each agent auto-discovers (.claude/skills/ and .codex/skills/), and every tool picks them up automatically.

Available Skills

Skill Purpose
caveman Ultra-compressed communication mode, cuts token usage ~75%.
context-audit Audit your Claude Code setup for token waste and context bloat.
diagnose Disciplined diagnosis loop for hard bugs and performance regressions.
fewer-permission-prompts Add allow rules to reduce repetitive permission prompts.
find-skills Discover and install agent skills for a given task.
grill-me Interview-style stress-test of a plan or design.
grill-with-docs Grilling session that challenges your plan against the domain model and updates docs inline.
improve-codebase-architecture Find refactoring and architecture improvement opportunities.
init Initialize a CLAUDE.md for the current project.
jira-helper Create, draft, query, and summarize Jira Cloud tickets from conversations, notes, or files.
pm Parallelized work breakdown, effort estimates, and team-ready task list.
prompt-prep Prepare a compact, structured Agent task prompt from a small set of inputs.
review Review a pull request or branch changes for bugs and improvements.
security-review Security-focused review of changed code. Checks for OWASP top 10.
setup-skills Scaffold per-repo config (issue tracker, triage labels, domain docs) for engineering skills.
simplify Review changed code for unnecessary complexity and fix it.
tdd Test-driven development with red-green-refactor loop.
to-issues Break a plan or PRD into independently-grabbable issues on the issue tracker.
to-prd Turn the current conversation context into a PRD on the issue tracker.
triage Move issues through a triage state machine.
write-a-skill Create new agent skills with proper structure.
zoom-out Step back and assess the bigger picture of a codebase or problem.

Install

Two ways to install — identical result. Pick whichever fits your team.

uv (recommended — no Python required)

curl -LsSf https://astral.sh/uv/install.sh | sh   # once, if you don't have uv
uvx agent-skills-bundle install review tdd                  # named skills
uvx agent-skills-bundle install --all                       # everything
uvx agent-skills-bundle==0.2.0 install review               # pin a version

bash (local clone — no tooling required)

git clone https://github.com/priyankaiiit14/agents-skills.git
cd agents-skills
./scripts/install-local.sh /path/to/project        # omit the path to use the current dir

Both accept the --target and --scope flags described next.

Targets and scopes

--target chooses the agent(s); --scope chooses project vs machine-wide. Skills are copied to the matching directory, which each agent auto-discovers at startup:

--target claude --target codex
--scope project (default) <project>/.claude/skills/ <project>/.codex/skills/
--scope global ~/.claude/skills/ ~/.codex/skills/

Defaults are --target both --scope project. Use project scope to commit skills to a repo so teammates get them on git pull; use global scope to make them available in every project on your machine.

uvx agent-skills-bundle install review                    # both agents, project (default)
uvx agent-skills-bundle install review --target codex     # Codex only
uvx agent-skills-bundle install --all --scope global      # every skill, machine-wide

A lockfile records what's installed and the target/scope used — skills-lock.json (project) or ~/.agent-skills-bundle/skills-lock.json (global).

Commands

Command What it does
agent-skills-bundle list List every skill in the installed version, with descriptions
agent-skills-bundle install <names…> Install named skills (--all for everything)
agent-skills-bundle update Re-install tracked skills at the current version
agent-skills-bundle status Show which skills are installed vs available, and their versions
agent-skills-bundle create <name> Scaffold a new skill folder (run from a repo clone)

install, update, and status accept --target and --scope; update reuses the target/scope saved in the lockfile. Run any command with --help for details.

Publishing to PyPI

One-time setup

Done once, before the first release. Publishing uses PyPI Trusted Publishing (OIDC) — no API tokens are stored anywhere.

  1. Create a PyPI account.
  2. Confirm the project name agent-skills-bundle is free on PyPI (rename it in pyproject.toml if taken).
  3. On PyPI, add a Trusted Publisher (Account settings → Publishing) with:
    • Repository: priyankaiiit14/agents-skills
    • Workflow: publish.yml
    • Environment: pypi
  4. In the GitHub repo, create an environment named pypi (Settings → Environments).

Testing locally before publishing

After making changes, verify the package works before opening a PR:

Option 1 — editable install (fastest for iteration)

pip install -e .
agent-skills-bundle list
agent-skills-bundle install review --scope global

Changes to Python files take effect immediately; no reinstall needed.

Option 2 — install from a built wheel (closest to what PyPI ships)

uv build                                        # produces dist/agent_skills_bundle-<ver>-py3-none-any.whl
pip install dist/agent_skills_bundle-*.whl      # or: pip install --force-reinstall dist/...whl
agent-skills-bundle list

Use this when you want to confirm the packaged artefact (not just the source) works end-to-end.

Option 3 — run directly without installing

uvx --from . agent-skills-bundle list

Useful for a quick smoke-test in CI or a clean shell where you don't want to touch the system site-packages.

Cutting a release

Bump the version in your PR, then merge to main — publishing is automatic.

  1. Bump version in pyproject.toml and src/agent_skills_bundle/__init__.py (keep them in sync).
  2. Open a PR, get it reviewed, and merge to main.

The GitHub Action (.github/workflows/publish.yml) runs on every push to main. It checks whether the current package version is already on PyPI — if it is, the publish step is skipped; if it is new, the package is built and published automatically. Teams then pin with uvx agent-skills-bundle==0.2.0 install or run uvx agent-skills-bundle update to pick up the new version.

Contributing a skill

You do not need to clone the repo to use skills. Clone only to add or edit one. All skills live in src/agent_skills_bundle/skills/ — that's the single source of truth, and both agents' copies are generated from it.

git clone git@github.com:priyankaiiit14/agents-skills.git
cd agents-skills

Add a new skill

git checkout -b skills/my-skill-name
uvx agent-skills-bundle create my-skill-name   # scaffolds src/agent_skills_bundle/skills/my-skill-name/SKILL.md
# edit SKILL.md, then:
git add src/agent_skills_bundle/skills/my-skill-name
git commit -m "add my-skill-name skill"
# open a PR to main

create writes the folder with the required frontmatter already filled in. Then add the skill to the table at the top of this README.

Update an existing skill

git checkout -b skills/my-skill-name        # or fix/my-skill-name for a bug fix
# edit src/agent_skills_bundle/skills/my-skill-name/SKILL.md (or its references/, scripts/, assets/)
git add src/agent_skills_bundle/skills/my-skill-name
git commit -m "update my-skill-name: <what changed>"
# open a PR to main

The change ships to users in the next release (see How releases work). Users pull it with uvx agent-skills-bundle update.

Branch naming

Branch Purpose
main Always stable. Protected — PRs required. Merging here triggers auto-publish if the version changed.
skills/<name> New skill or feature change to an existing one
fix/<name> Bug fix to an existing skill
feature/<name> Version bump + any other non-skill change

Skill layout

src/agent_skills_bundle/skills/
  <skill-name>/
    SKILL.md          ← required; keep concise
    references/       ← large or conditional reference material
    scripts/          ← executable helpers
    assets/           ← templates, binaries, examples

Every SKILL.md needs YAML frontmatter — name and description are required (the description is what shows in agent-skills-bundle list and what each agent uses to decide when to trigger the skill):

---
name: my-skill-name
description: One-line description of what this skill does and when to invoke it.
---

Download files

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

Source Distribution

agent_skills_bundle-0.1.0.tar.gz (70.3 kB view details)

Uploaded Source

Built Distribution

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

agent_skills_bundle-0.1.0-py3-none-any.whl (84.1 kB view details)

Uploaded Python 3

File details

Details for the file agent_skills_bundle-0.1.0.tar.gz.

File metadata

  • Download URL: agent_skills_bundle-0.1.0.tar.gz
  • Upload date:
  • Size: 70.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.14

File hashes

Hashes for agent_skills_bundle-0.1.0.tar.gz
Algorithm Hash digest
SHA256 1df393b225590d813e2fa6f6e658de89d1f352cb9ffd126ebd0f3a275980bad5
MD5 160a5384aa281836f2ed170d2a05bf47
BLAKE2b-256 7a9a6b5a6ad35a1750c1cdbb635a0f1e038346766158afbdbda8aeaa39841e81

See more details on using hashes here.

Provenance

The following attestation bundles were made for agent_skills_bundle-0.1.0.tar.gz:

Publisher: publish.yml on priyankaiiit14/agents-skills

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

File details

Details for the file agent_skills_bundle-0.1.0-py3-none-any.whl.

File metadata

File hashes

Hashes for agent_skills_bundle-0.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 c451e2e2ffbacbbcc679e9341b89754a5bffc2a384602ecb19b00d64bab5f52d
MD5 ed3da73b09e2ab23bc264a9051d8a52e
BLAKE2b-256 71f6925c4b1ba4ee31813b6e85758ebf86cceb4505d50d57e298479cc9b5a78d

See more details on using hashes here.

Provenance

The following attestation bundles were made for agent_skills_bundle-0.1.0-py3-none-any.whl:

Publisher: publish.yml on priyankaiiit14/agents-skills

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 Sentry Error logging StatusPage Status page