SUPER_COMPUTER_AI
AGENTS.md0 starsChanged 47 days agoMITView on GitHub
This repository is the single home for AI agent skills, agent roles, instructions and MCP servers used with Claude Code and Codex. CLAUDE.md imports this file; do not duplicate rules there. Names are kebab-case. Skill names are globally unique regardless of category. - Never commit secrets, tokens, API keys, .env files, account/customer IDs, e-mail addresses, browser profiles, or machine-specific absolute paths (/Users/..., /home/...). - Credentials live in ~/.config/super-computer-ai/credentials/<server>/ (dir 0700, files 0600). Sensitive project values live in ~/.config/super-computer-ai/private/. Commit only…
- Reads credentials
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 31,774 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.
- CLAUDE.mdCLAUDE.md
- google-ads-managementSkill
- meta-ads-managementSkill
- tiktok-ads-managementSkill
- ai-automationSkill
- ai-researchSkill
- ai-system-designSkill
- context-engineeringSkill
- eval-harnessSkill
- implementation-verificationSkill
- inference-optimizationSkill
- model-selection-hfSkill
- model-servingSkill
- production-opsSkill
- rag-engineeringSkill
- tool-and-mcp-designSkill
- training-data-engineeringSkill
- training-peftSkill
- jev-browserSkill
- premiere-pro-mcpSkill
- suno-music-generatorSkill
- growth-loopSkill
- mobile-store-releaseSkill
- mobile-ui-reviewSkill
- store-listing-assetsSkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are 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.
3
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 21
Never commit secrets, tokens, API keys, .env files, account/customer IDs, e-mail addresses, browser profiles, or machine-specific absolute …
- Very specific: line 27
Run python3 scripts/validate.py and python3 scripts/skills/validate_skills.py before every commit.
- Very specific: line 32
Never hand-edit ~/.claude, ~/.codex or ~/.agents.
- Distinctive: line 40
Do not initialize, commit, push or rewrite Git history without explicit human approval.
≈ “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
SUPER_COMPUTER_AI — Working Rules
This repository is the single home for AI agent skills, agent roles, instructions and MCP servers used with
Claude Code and Codex. CLAUDE.md imports this file; do not duplicate rules there.
Where things go
| You are adding | Put it in | Also update |
| A skill | skills/<category>/<skill-name>/SKILL.md (+ references/, agents/openai.yaml) | skills/manifest.json |
| An agent role | agents/claude/<role>.md and agents/codex/<role>.toml | agents/README.md |
| An instruction | instructions/global/ (all sessions), instructions/pipeline/, or instructions/projects/ | — |
| An MCP server (our code) | mcp/servers/<name>/ with its own README, .env.example, tests and lockfile | mcp/registry.json |
| A hosted MCP (not our code) | mcp/remote/<name>/ (README + auth helper if needed) | mcp/registry.json |
| A helper tool | tools/<name>/ | — |
Names are kebab-case. Skill names are globally unique regardless of category.
Public-repo rules (non-negotiable)
- Never commit secrets, tokens, API keys,
.envfiles, account/customer IDs, e-mail addresses, browser profiles, or machine-specific absolute paths (/Users/...,/home/...). - Credentials live in
~/.config/super-computer-ai/credentials/<server>/(dir0700, files0600). Sensitive project values live in~/.config/super-computer-ai/private/. Commit only variable names and.env.examplefiles. - Migrations from other repositories copy only files on an explicit, reviewed allowlist.
- Run
python3 scripts/validate.pyandpython3 scripts/skills/validate_skills.pybefore every commit; CI runs both plus a gitleaks scan.
Installing into Claude Code / Codex
Never hand-edit ~/.claude, ~/.codex or ~/.agents. The installer (scripts/install.py, see docs/installing.md) links
skills and roles and writes the managed instruction block. Projects outside this repo's scope
(e.g. AI_ORCHESTRATION) are never touched by the installer.
Working agreements
- Plans and design decisions are written to files; implementation follows an approved plan.
- Coder and reviewer roles are assigned per task by the human; the coder sends changes for peer review.
- Do not initialize, commit, push or rewrite Git history without explicit human approval.
- Evidence before assertions: run the deliverable, record the command and result, mark anything not
exercised as
UNVERIFIED.
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.