dazbo-agent-skills
AGENTS.md28 starsChanged yesterdayMITView on GitHub
Public collection of portable, self-contained AI agent skills authored by Dazbo (derailed-dash) and distributed via Skills.sh (npx skills add https://github.com/derailed-dash/dazbo-agent-skills).
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 32,671 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository10 other files it gives its agents.
The file
We split this AGENTS.md into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 16
Plain-English Opening: Start every SKILL.md with # Heading 1 followed immediately by 1–2 plain-English paragraphs explaining *what* the …
- Distinctive: line 34
Run the offline static and unit-test gate from the repository root before committing:
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Repository Guidelines (dazbo-agent-skills)
Public collection of portable, self-contained AI agent skills authored by Dazbo (derailed-dash) and distributed via Skills.sh (npx skills add https://github.com/derailed-dash/dazbo-agent-skills).
Key References
- Storefront & Catalogue: README.md (
Available Skillstable,npxinstall commands, project tree, and companion articles) - Skill Entrypoints:
skills/<skill-name>/SKILL.md(with optionalscripts/andreferences/subdirectories) - Test & Lint Harness: Makefile, pyproject.toml, and
tests/unit/ - CI & Automated Review: .github/workflows/pr-checks.yml, .github/workflows/gemini-review.yml, and .github/workflows/gemini-triage.yml
- Upstream Format: Agent Skills Open Specification (
agentskills.io)
Skill Authoring Principles
- Naming & Frontmatter: Use clear kebab-case folder names (preferring
verb(ing)-nounor explicit action-noun phrasing). EverySKILL.mdmust begin with YAML frontmatter wherenamematches the folder name,descriptionis written in the 3rd person with explicit"Use when..."activation triggers (and"Don't use..."boundaries when helpful), andmetadatasetsauthor: Dazbo (GitHub: derailed-dash)andrepository: https://github.com/derailed-dash/dazbo-agent-skills(plus optionalclidependencies). - Plain-English Opening: Start every
SKILL.mdwith# Heading 1followed immediately by 1–2 plain-English paragraphs explaining what the skill does and how it works conceptually before introducing dense tables, flags, or rubrics. - Progressive Disclosure: Keep
SKILL.mdunder 500 lines. Move templates, samples, and deep reference tables intoreferences/(linked via Unix-style relative paths for on-demand reading), and place reusable deterministic utilities intoscripts/(documenting expected arguments, prerequisites, and output format). Add a## Table of ContentswheneverSKILL.mdor a reference Markdown file exceeds 100 lines. - High-Signal Guidance: Assume a capable frontier model — omit generic explanations the model already knows. Prefer numbered workflows with explicit intermediate validation steps, deterministic commands for low-freedom tasks, and a single recommended path rather than competing options.
- Portable Script & Tool References: Reference bundled scripts via installed skill paths (
~/.gemini/config/skills/<skill-name>/scripts/...and.agents/skills/<skill-name>/scripts/...) rather than assuming execution from a clone of this repository. When referencing MCP tools, identify both the MCP server and tool name clearly rather than bare tool names.
Repository Invariants (B1–B8)
- B1
[det]: Everyskills/<skill-name>/SKILL.mdstays strictly under500lines, has YAMLnamematching<skill-name>and a 3rd-person"Use when..."description, and includes## Table of Contentswhenever over100lines. - B2
[det]: Every Markdown heading (#–####) is followed immediately by a blank line, fenced code blocks close cleanly (use~~~~outer fences when nesting triple-backtick examples), and internal#anchorlinks resolve to valid GitHub-flavoured slugs. - B3
[det]: All prose, documentation, and comments use English (UK) spelling (e.g.synchronise,organise,artefacts,initialisation,summarise) and passuv run codespell -s. - B4
[det]: Markdown files use relative links only (neverfile:///URIs) and wrap literal dollar amounts in inline backticks (e.g.`$0`) rather than backslash-escaping\$0. - B5
[det]: Zeromermaidblocks or ASCII/Unicode box-drawing diagrams appear in any file (generate PNG/SVG image assets when architecture or workflow diagrams are needed). - B6
[det]: Python utilities underskills/*/scripts/use Python>= 3.12modern type hints (list[str],str | None), include a top-of-module docstring, are registered in[tool.ruff]and[tool.pyright]in pyproject.toml with unit tests undertests/unit/, and passmake check. - B7: Adding, renaming, or materially changing a skill updates README.md (
Available Skillstable,npx skills addlist, andProject Structuretree) in the same change. - B8: Public-repository hygiene — never commit credentials, API keys, personal cloud project IDs, or machine-private local paths.
Verification
Run the offline static and unit-test gate from the repository root before committing:
make check # codespell, ruff check, ruff format --check, pyright, and pytest
make fix # auto-apply codespell and ruff fixesDoes it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

