googlecloud-plugin
CLAUDE.md4 starsChanged 4 months agoMITView on GitHub
This file advises AI agents working ON the plugin (building, researching, validating skills). It is committed, and public — write it for contributors, not as a scratchpad. For local notes you don't want published, use .claude/local-CLAUDE.md, which is gitignored. You are building a public plugin that installs a complete GCP team for other developers. Every file you commit must be production quality — real developers will use this. Never commit _plugin/ content. Draft in _plugin/staging/, promote to skills/ only after…
In this repository21 other files it gives its agents.
- AGENTS.mdAGENTS.md
- agent-architectSkill
- bigquerySkill
- cloud-runSkill
- cloud-storageSkill
- gcloud-guardrailsSkill
- gcp-architectSkill
- gcp-opsSkill
- gcp-project-contextSkill
- gcp-qaSkill
- gcp-securitySkill
- gkeSkill
- google-skills-catalogSkill
- iamSkill
- logging-monitoringSkill
- mcp-serversSkill
- networkingSkill
- solution-designerSkill
- terraform-gcpSkill
- vertex-aiSkill
- well-architectedSkill
The file
Show its rules and text
We split this CLAUDE.md into the rules it gives an agent
13 rules. About 0 are common, 6 are distinctive, and 7 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
7
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 10
Every file you commit must be production quality — real developers will use this.
- Very specific: line 19
Never commit _plugin/ content.
- Very specific: line 23
All work on feature branches: feat/phase-N-description
- Distinctive: line 25
Commit after each phase with a clear message
- Distinctive: line 29
Before any skill merges to main:
- Very specific: line 52
Never reintroduce a top-level triggers, version or required_scopes field — that is the exact regression make validate exists to catch.
- Very specific: line 56
Run make spec && make validate after any SKILL.md or manifest edit.
- Distinctive: line 60
No credential values in any file — location only
- Distinctive: line 62
No public Cloud Storage buckets without explicit design intent
- Distinctive: line 63
Every billable action gets a cost warning comment before the command
- Very specific: line 68
Run make crawl to refresh research/raw/ and research/SOURCES.md.
- Very specific: line 69
Run make check to detect URL drift.
- Very specific: line 80
Update research/coverage-matrix.md whenever a skill status changes: missing → researched → drafted → normalized → validated
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.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
GoogleCloud Plugin — Construction Guide
This file advises AI agents working ON the plugin (building, researching, validating skills).
It is committed, and public — write it for contributors, not as a scratchpad.
For local notes you don't want published, use .claude/local-CLAUDE.md, which is gitignored.
Your Role
You are building a public plugin that installs a complete GCP team for other developers. Every file you commit must be production quality — real developers will use this.
The Split
| Location | Purpose | Committed? |
googlecloud-plugin/ | Public release — curated, validated | Yes (main is protected) |
_plugin/ | Working area — planning, research drafts, audits | No (gitignored) |
Never commit _plugin/ content. Draft in _plugin/staging/, promote to skills/ only after make gate passes.
Branch Strategy
- All work on feature branches:
feat/phase-N-description - Main is branch-protected — PRs required
- Commit after each phase with a clear message
Gate System (Sequential — Enforce This)
Before any skill merges to main:
make gatepasses (validate + lint + test)- PR reviewed (devex-charter pr-reviewer or adversarial-gate for high-stakes changes)
- Coverage matrix updated in
research/coverage-matrix.md
Skill Contract
Frontmatter follows the Agent Skills specification.
The field set is closed — adding a top-level key outside
{name, description, license, compatibility, allowed-tools, metadata} makes the
skill non-conformant, and a conformant client MUST skip it. Everything specific
to this plugin is namespaced under metadata as string values:
name— matches directory name exactlydescription— what it does AND when to use it, max 1024 chars, ending withUse when the user mentions: …. This is the only thing a conformant client routes on. Where two skills overlap, name the other and draw the line.metadata["googlecloud-plugin/version"]metadata["googlecloud-plugin/triggers"]— comma-joined, non-emptymetadata["googlecloud-plugin/required-scopes"]— comma-joined,""if nonemetadata["googlecloud-plugin/mcp-servers"]— comma-joined, omit if none- persona skills add
persona,tier,gateunder the same namespace
Never reintroduce a top-level triggers, version or required_scopes field —
that is the exact regression make validate exists to catch. See
ADR-007.
Run make spec && make validate after any SKILL.md or manifest edit.
Safety Rules (Non-Negotiable)
- No credential values in any file — location only
- No
--allow-unauthenticatedCloud Run without gcp-security sign-off in HLD - No public Cloud Storage buckets without explicit design intent
- Every billable action gets a cost warning comment before the command
- IAM: never grant
roles/ownerorroles/editorin skill examples
Research Pipeline
Run make crawl to refresh research/raw/ and research/SOURCES.md.
Run make check to detect URL drift.
Research output lives in _plugin/research/ (gitignored) before being promoted to research/raw/.
Architecture Docs
- HLD for every feature touching ≥2 components →
architecture/HLD/ - LLD where HLD reveals complexity →
architecture/LLD/ - ADRs for significant decisions →
architecture/decisions/ADR-NNN-title.md
Coverage Matrix
Update research/coverage-matrix.md whenever a skill status changes:
missing → researched → drafted → normalized → validated
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.

