alps
AGENTS.md1 starsChanged 23 days agoApache-2.0View on GitHub
Japanese translation These instructions apply to the repository. Follow the user's authorized scope and preserve unrelated work and user-authored data. localization.yaml defines English as the source locale and Japanese as supported. Do not infer Process requirements from tests, templates, Host manifests, icons, or other presentation resources. Read the complete SKILL.md for each Skill selected before applying it. skills/ is the source of Plugin Skills. Hosts discover it through their applicable conventions and manifests. .agents/skills/ is an integrated repository view, not…
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 repository5 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- review-alpsSkill
- sync-localesSkill
- design-agent-work-systemSkill
- design-process-descriptionSkill
The file
We split this AGENTS.md into the rules it gives an agent
8 rules. About 0 are common, 2 are distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 18
Do not infer Process requirements from tests, templates, Host manifests, icons, or other presentation resources.
- Very specific: line 40
Use review-alps for changes to the Process Framework or work-system design principles, Skill content, repository guidance, tests, …
- Very specific: line 41
Use sync-locales for each affected English/Japanese pair.
- Very specific: line 42
Use design-process-description when authoring or reviewing a Process Description.
- Very specific: line 43
Use design-agent-work-system when designing or reviewing the supporting configuration, implementation, or effectiveness.
- Distinctive: line 44
Keep each Host adapter aligned with its native format and the distribution layout above.
- Distinctive: line 49
Keep the following evidence distinct:
- Very specific: line 58
Run the checks in .github/workflows/validate.yml that the environment permits.
≈ “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 Instructions
Japanese translation
These instructions apply to the repository. Follow the user's authorized scope and preserve unrelated work and user-authored data. localization.yaml defines English as the source locale and Japanese as supported.
Sources
| Subject | Source |
| Process meaning, boundaries, references, change, and evaluation | Process Framework |
| Agent, tool, information, and environment design | Design Principles for Agent Work Systems |
Writing Skill bodies: the Process Description in each SKILL.md | Process Framework |
| Design work and its application to Agent Skills | design-process-description and design-agent-work-system |
| Repository work and distribution | This file |
| Drafting aids | Template and examples, both informative |
Do not infer Process requirements from tests, templates, Host manifests, icons, or other presentation resources. Read the complete SKILL.md for each Skill selected before applying it.
Layout and distribution
| Path | Role |
skills/design-process-description/, skills/design-agent-work-system/ | Distributed Skills; each root English SKILL.md is the source for its Process. |
.agents/skills/<distributed-skill> | Relative symlink to ../../skills/<distributed-skill> for repository discovery. |
.agents/skills/review-alps/ | Real directory for repository semantic and distribution review; not a Plugin Skill. |
.agents/skills/sync-locales/ | Real directory for English/Japanese review; not a Plugin Skill. |
Each Skill's references/ | Principles and supporting material used by that Skill; each document's role determines its normative force. |
docs/locales/ja/ and each Skill's references/locales/ja/ | Supported translations of the corresponding English source files. A translated Skill body is named SKILL.ja.md and is not a Skill entry point. |
examples/ | Bundled reference material, including a working target Skill; outside Plugin Skill discovery. The guide's translation is in examples/locales/ja/, and the target Skill's translations are in its references/locales/ja/. |
plugin.json, .claude-plugin/, .cursor-plugin/, .codex-plugin/ | Root Plugin format and distinct Host adapters. |
CLAUDE.md | Claude Code entry point for repository development that imports this file; it adds no instructions of its own and is not Plugin context. |
assets/ and the Skill's agents/ and assets/ | Presentation resources. |
skills/ is the source of Plugin Skills. Hosts discover it through their applicable conventions and manifests. .agents/skills/ is an integrated repository view, not a universal Host convention. A checkout can contain development Skills without exposing them as Plugin Skills. Each distributed Skill functions on its own: its relative links resolve within its own directory, and it does not depend on the other Skill by file or name. When the Skills are combined, they exchange only a description of the work. The Plugin distributes both Skills together with examples/ as reference material; the Skills do not link to it. Keep development Skills and example Skills outside skills/.
Change and review
- Inspect current files and the task-owned diff before editing. Preserve unrelated work.
- Use
review-alpsfor changes to the Process Framework or work-system design principles, Skill content, repository guidance, tests, distribution, or presentation that affect their meaning or boundaries. - Use
sync-localesfor each affected English/Japanese pair. Repository-development Skills have no Japanese Plugin counterparts. - Use
design-process-descriptionwhen authoring or reviewing a Process Description. - Use
design-agent-work-systemwhen designing or reviewing the supporting configuration, implementation, or effectiveness. Preserve shared work meaning and apply each foundation within its subject. - Keep each Host adapter aligned with its native format and the distribution layout above.
- Apply CONTRIBUTING for contribution and licensing requirements and Versioning for release policy.
Verification and delivery
Keep the following evidence distinct:
- Agent Skill and Plugin form validation against their applicable formats.
- Repository validation: automated checks of Plugin manifest consistency, distribution layout, translation manifests, and relative links, and review of affected file layout, symlinks, Host resources, and distribution boundaries. After reviewing a Japanese counterpart against its English source, update its
reviewed_source_sha256in the translation manifest of itslocales/ja/directory. - Semantic review: purpose and Outcome sufficiency, required details and obligations, references, evaluation limits, and English/Japanese meaning and normative force.
- Tool and connection verification against the specified behavior, including failures and incomplete results.
- Respect for applicable requirements, authority, and limits on effects, including those implemented by tools, assessed separately from tool behavior and effectiveness.
- Work-system effectiveness through representative work with an agent, information, tools, and environment, assessed against the work's Outcomes and conditions.
Run the checks in .github/workflows/validate.yml that the environment permits. At minimum run python3 -m unittest discover -s tests -v, git diff --check, changed-link checks, and review the complete task-owned diff, including new files. Passing mechanical checks does not prove semantic validity or successful Process execution.
Report findings, completed checks, failed or unperformed checks, and limits.
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.

