agent-guidance-kit
AGENTS.md1 starsChanged 15 days agoApache-2.0View on GitHub
This repository is a portable library of agent skills. When working here: - The canonical catalog lives in skills/; each skill is a self-contained SKILL.md following the Agent Skills format. The README's "Included Skills" table is the index; the full adoption workflow is in docs/using-the-library.md. - Keep skills portable: use project-neutral names, state triggers and boundaries, and put deeper material in skill-local references/ files. - Do not modify ordinary skill bodies or references for cleanup, style, or benchmark results. Skill…
- Installs packages
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 31,843 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 repository25 other files it gives its agents.
- adversarial-pr-reviewSkill
- ai-slop-detectorSkill
- architecture-reviewSkill
- codebase-orientationSkill
- code-reviewSkill
- dependency-upgradeSkill
- documentation-reviewSkill
- frontend-quality-reviewSkill
- git-github-workflowSkill
- harness-adaptationSkill
- implementation-planningSkill
- parallel-multi-agentSkill
- quality-hardeningSkill
- reduce-code-sizeSkill
- repository-guidance-authoringSkill
- requirements-and-designSkill
- review-feedback-resolutionSkill
- rules-and-skills-auditSkill
- security-reviewSkill
- skill-authoringSkill
- skill-discoverySkill
- skill-optimizerSkill
- skill-reviewerSkill
- systematic-debuggingSkill
- threat-modelingSkill
The file
We split this AGENTS.md into the rules it gives an agent
6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
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
- Very specific: line 9
Keep skills portable: use project-neutral names, state triggers and boundaries, and put deeper material in skill-local references/ files.
- Distinctive: line 11
Do not modify ordinary skill bodies or references for cleanup, style, or benchmark results.
- Distinctive: line 22
Run the product-only gate before committing and again before pushing:
- Distinctive: line 39
Every branch that opens or updates a pull request MUST receive a completed fresh-context adversarial review before any push.
- Distinctive: line 41
After a PR is created, review the complete update delta before every later push until the review converges with no additional findings.
- Distinctive: line 49
If a fresh read-only reviewer cannot be launched, the review is blocked: do not push and report the exact capability gap.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 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
Agent Guidance Kit — Agent Instructions
This repository is a portable library of agent skills. When working here:
- The canonical catalog lives in
skills/; each skill is a self-containedSKILL.mdfollowing the Agent Skills format. The README's "Included Skills" table is the index; the full adoption workflow is indocs/using-the-library.md. - Keep skills portable: use project-neutral names, state triggers and
boundaries, and put deeper material in skill-local
references/files. - Do not modify ordinary skill bodies or references for cleanup, style, or benchmark results. Skill changes require an explicitly approved change pass.
Deterministic Gate
Install the pinned validation dependencies before running the gate:
python3 -m pip install -r requirements-dev.txtRun the product-only gate before committing and again before pushing:
python3 scripts/validate_catalog.py
python3 -m unittest discover -s scripts -p 'test_*.py' -q
ruff check scripts/
git diff --checkThe gate checks skill structure and frontmatter, README catalog integrity, and repository-relative Markdown links.
Mandatory Adversarial Review for PRs
Per repository policy and
skills/adversarial-pr-review/SKILL.md:
- Every branch that opens or updates a pull request MUST receive a completed fresh-context adversarial review before any push.
- After a PR is created, review the complete update delta before every later push until the review converges with no additional findings.
- The parent may not substitute its own self-review; at least one fresh, independent, read-only subagent reviewer is required.
- Review findings are applied only in a separately authorized change pass.
- Record a compact
PASSor up to three anchored findings with the reviewed commit range.
If a fresh read-only reviewer cannot be launched, the review is blocked: do not push and report the exact capability gap.
Does 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.

