fulcrum / rules
AGENTS.md4 starsChanged 7 months agoMITView on GitHub
Fulcrum installs this file once under ~/.fulcrum/rules/AGENTS.md. Agent rule files must import it or point to it. Do not paste this body into every agent file. Before suggesting commands, editing files, or running tools in any project, read available project guidance: Project guidance wins over defaults in this file and linked situational rules. When a trigger applies, read the linked file before acting. If this file is installed at ~/.fulcrum/rules/AGENTS.md, resolve links under ~/.fulcrum/rules/. Fulcrum owns only the small loader…
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 repository35 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/agents.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- gh-authoredSkill
- batSkill
- biomeSkill
- dart-toolchainSkill
- difftasticSkill
- direnvSkill
- ezaSkill
- flarectlSkill
- fzfSkill
- ghSkill
- git-cliffSkill
- gitleaksSkill
- google-java-formatSkill
- hyperfineSkill
- jqSkill
- justSkill
- ktlintSkill
- lizardSkill
- macro-subagent-orchestrationSkill
- miseSkill
- osv-scannerSkill
- pmdSkill
- ruffSkill
- sdSkill
- spotbugsSkill
- subagent-orchestrationSkill
- tool-nameSkill
- usqlSkill
- watchexecSkill
- xhSkill
- yqSkill
- zoxideSkill
The file
We split this AGENTS.md into the rules it gives an agent
10 rules. About 0 are common, 6 are distinctive, and 4 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.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
Before suggesting commands, editing files, or running tools in any project, read available project guidance:
- Very specific: line 28
Use machine-readable CLI output when available: --json, --format json, -o json, then parse with jq or yq.
- Distinctive: line 29
Do not dump large raw output into chat.
- Very specific: line 30
Do not write to ~/.agents/.
- Distinctive: line 32
Ask before destructive actions: recursive delete, hard reset, force push, git clean, database drop or truncate, infrastructure destroy, …
- Distinctive: line 33
Verify before claiming done: run the project runner, lint, typecheck, tests, build, or a targeted smoke check.
- Distinctive: line 34
Keep code names based on responsibility, domain, behavior, or value.
- Distinctive: line 35
Prefer editing existing modules over adding overlapping utilities.
- Very specific: line 40
When a trigger applies, read the linked file before acting.
- Very specific: line 43
Search, data tools, HTTP, browser, database, performance, and process handling: situational/tooling.md
≈ “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
Fulcrum rule index
Fulcrum installs this file once under ~/.fulcrum/rules/AGENTS.md. Agent rule files must import it or point to it. Do not paste this body into every agent file.
Install contract
- Agents with native
@import support load this file by import. - Agents without native import support receive a small loader that forces a read of this file before planning, editing, or running commands.
- Detailed rules live in linked markdown files next to this file. Read them only when their trigger applies.
- User content outside
<!-- BEGIN/END FULCRUM RULES -->markers in agent files is preserved.
Must read first
Before suggesting commands, editing files, or running tools in any project, read available project guidance:
AGENTS.mdCLAUDE.mdjustfile.mise.tomlor.tool-versions.envrcpyproject.tomlpackage.json
Project guidance wins over defaults in this file and linked situational rules.
Always-on rules
- Use machine-readable CLI output when available:
--json,--format json,-o json, then parse withjqoryq. - Do not dump large raw output into chat. Use indexed, batched, or sandboxed processing and return the answer or artifact path.
- Do not write to
~/.agents/. Use each agent's native skill or rules directory. - Preserve user content outside Fulcrum sentinels.
- Ask before destructive actions: recursive delete, hard reset, force push, git clean, database drop or truncate, infrastructure destroy, production delete.
- Verify before claiming done: run the project runner, lint, typecheck, tests, build, or a targeted smoke check.
- Keep code names based on responsibility, domain, behavior, or value. Do not name code after plan phase, provenance, status, or inspiration source.
- Prefer editing existing modules over adding overlapping utilities.
- Comment why a non-obvious choice exists. Do not comment what the next line already says.
Situational reads
When a trigger applies, read the linked file before acting. If this file is installed at ~/.fulcrum/rules/AGENTS.md, resolve links under ~/.fulcrum/rules/.
- Project discovery, runners, package managers, environment setup:
situational/project-discovery.md - Search, data tools, HTTP, browser, database, performance, and process handling:
situational/tooling.md - Git workflow, diffs, commits, verification, and security scans:
situational/git-quality-security.md - Architecture, naming, test scope, and persistence or server stack choices:
situational/architecture-and-code.md - Context-mode routing, subagents, output style, and long-session context control:
situational/orchestration-and-context.md
Vendor rule note
Fulcrum owns only the small loader block in each agent's primary rules file. Vendor hooks, plugins, settings, and skills stay in their native locations.
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.

