antigravity-harness / rules
AGENTS.md1 starsChanged 17 days agoApache-2.0View on GitHub
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 repository7 other files it gives its agents.
- harness-controlSkill
- harness-handoffSkill
- harness-recallSkill
- harness-reflectorSkill
- harness-skill-authorSkill
- systematic-debuggingSkill
- verification-before-completionSkill
The file
We split this AGENTS.md into the rules it gives an agent
9 rules. About 1 is common, 7 are distinctive, and 1 is very specific to this project.
≈1
Common · estimate
Many files like this have a rule of this kind.
≈7
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
- Distinctive: line 7
Use existing learned skills before re-deriving a known multi-step procedure.
- Very specific: line 8
Use the harness CLI (bin/harness.cmd on Windows, bin/harness elsewhere) to inspect memory and validate skills.
- Distinctive: line 9
Keep the Harness-managed AGENTS.md block below 20 KB.
- Common: line 10
Never persist credentials, tokens, private keys, or connection strings in rules, skills, or memory exports.
- Distinctive: line 16
When a command fails, capture the full stderr and the first actionable error line.
- Distinctive: line 19
Write and read text files as UTF-8 to avoid cp1252 corruption on Windows.
- Distinctive: line 20
Do not copy project-specific commands or paths into the global rules file.
- Distinctive: line 21
Only promote a lesson after observing a successful recovery.
- Distinctive: line 22
Before overwriting a skill, check whether it is manually authored.
≈ “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
<!-- HARNESS:AUTONOMOUS_RULES_START -->Harness Evolved Directives (Global Environment)
Maintained automatically by Harness². User content outside this block is preserved. Verified failure/recovery lessons only; no transcript timestamps or exit-code noise.
Standing Project Directives
- Use existing learned skills before re-deriving a known multi-step procedure.
- Use the harness CLI (
bin/harness.cmdon Windows,bin/harnesselsewhere) to inspect memory and validate skills. - Keep the Harness-managed
AGENTS.mdblock below 20 KB; Antigravity caps each rules file at 24 KB. - Never persist credentials, tokens, private keys, or connection strings in rules, skills, or memory exports.
Environment & Tooling Quirks
- PowerShell: quote paths containing spaces and avoid multi-line inline commands; write complex logic to a temporary script.
- Python on Windows: run with
-X utf8when output contains non-ASCII, and preferpython -m moduleover bare entry points. - Antigravity hooks run with the plugin directory as the working directory; resolve project paths from the hook payload instead of the process cwd.
- When a command fails, capture the full stderr and the first actionable error line; an exit code alone is not a diagnosis.
Learned Pitfalls & Recovery Invariants
- Write and read text files as UTF-8 to avoid cp1252 corruption on Windows.
- Do not copy project-specific commands or paths into the global rules file; put them in the workspace rules instead.
- Only promote a lesson after observing a successful recovery; unresolved failures remain candidates.
- Before overwriting a skill, check whether it is manually authored; Harness must never replace user content.
<!-- HARNESS:AUTONOMOUS_RULES_END -->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.