instinct-help
Skill0 starsChanged 4 months agoMITView on GitHub
Use when the user asks what Instinct is, how it works, or wants a quick reference to the Gate/tiers without re-reading SKILL.md or the RFCs. Explicit invocation only.
In this repository8 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- .cursor/rules/instinct.mdcCursor rule
- instinct-auditSkill
- instinct-debtSkill
- instinct-reviewSkill
- instinctSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
No to all three → implement directly.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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
Instinct in one screen:
Identity: Engineering is not writing code. Engineering is making good decisions.
The Gate (three questions, stop at first yes, unclear counts as yes):
- Touches anything beyond what was named?
- More than one consumer?
- Expensive to undo (shipped, trust boundary, needs coordinated follow-up)?
No to all three → implement directly.
Tiers:
- Consequential (yes to Q1/Q2, no to Q3): look once for an existing pattern before writing new code, smallest diff, state why in the commit.
- Critical (yes to Q3): reconstruct why the current shape exists, state the strongest alternative once, implement whichever gets confirmed, stay scoped, deposit the reasoning somewhere durable.
Always: non-trivial logic leaves one runnable check behind. Bug fixes check for the same defect at other call sites before patching just the one path named. Never let judgment become nagging — one objection, stated once, then implement the decision.
Full detail: skills/instinct/SKILL.md (the actual behavior),
references/ladder.md (Critical-tier worked examples), docs/rfcs/ (why
every piece of this is shaped the way it is — 0001 for the philosophy,
0003 for the mechanism, 0007 for how this reaches hosts other than
Claude Code).
Related: instinct-review audits one already-made diff against this
same Gate; instinct-audit does the same at repo/history scale;
instinct-debt reports accepted instinct:-marked limitations. The core
instinct skill is ambient and applies before a change is made, not
after — the other three are all explicit-invocation, retrospective checks.
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.