principles
Skill41k starsChanged 7 days agoMITView on GitHub
Twenty-four engineering principles in five groups, mapped to the lane carrying each — premise testing, root causes, honest verification, domain modeling, convergence, context discipline. For design, debugging, or delegation briefs.
16 other projects use a version of this, 4 of them edited.
Trending
Quiet since January.thin dataabout the whole repository
Most recently taken up by
- scottlivingston/claude-skillstheir own version, first seen Nov 8, 2026
- Yeachan-Heo/oh-my-claudecodeword for word, first seen Nov 8, 2026
- endqwerty/endstackedited, first seen Nov 8, 2026
- Belweave/hstacktheir own version, first seen Nov 8, 2026
- Pixxle/skillsedited, first seen Nov 1, 2026
- christofferbergj/agent-skillstheir own version, first seen Nov 1, 2026
Who's talking about Yeachan-Heo/oh-my-claude-sisyphus
- A Hacker News story“Sisyphus Now Lives in Oh My Claude”Hacker News · Jan 11, 2026 · 52 points
- Yeachan-Heo/oh-my-claudecode“Silent update security issue”GitHub · Jan 10, 2026 · 1 reactions
- anthropics/claude-plugins-official“feat: Add oh-my-claude-sisyphus multi-agent orchestration plugin”GitHub · Jan 12, 2026
The biggest 3 of 3 mentions linking to the repository that our crawl found. Each line is their own headline.
Used with
Projects with it also use no-comments.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
jaydubya818Skillno-comments
Use when code comments need structural review, root-cause fixes, and constraint encoding.
Also in 5 projects that use this
Other versions16 projects use a version of this.
Word for word · 1
- Yeachan-Heo/oh-my-claudecode★ 41k100% same
Edited · 4
- overextended/skills★ 2223% same
- anis-dr/skills★ 321% same
- endqwerty/endstack★ 111% same
- Pixxle/skills★ 127% same
Wrote their own version · 11
- umararshad4/honey-stack★ 5same name
- stainless-code/lossless★ 4same name
- jonmumm/skills★ 2same name
- Autocss-com/ai★ 1same name
- Belweave/hstack★ 1same name
- christofferbergj/agent-skills★ 1same name
- pavelsimo/skills★ 1same name
- scottlivingston/claude-skills★ 1same name
- vasu-devs/Forge★ 1same name
- mukco/claude-skills★ 0same name
1 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository90 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- skills/AGENTS.mdAGENTS.md
- src/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- docs/CLAUDE.mdCLAUDE.md
- .github/CLAUDE.mdCLAUDE.md
- commandsSkill
- agent-doc-disciplineSkill
- ai-slop-cleanerSkill
- architectSkill
- architecture-surveySkill
- arenaSkill
- ask-navigatorSkill
- askSkill
- autopilotSkill
- autoresearchSkill
- babysitSkill
- benchmark-checklistSkill
- blast-radiusSkill
- broSkill
- cancelSkill
- configure-notificationsSkill
- correctSkill
- create-formal-verification-skillSkill
- create-verification-skillSkill
- debugSkill
- deepinitSkill
- deep-interviewSkill
- diagramSkill
- drydockSkill
- executeSkill
- external-contextSkill
- figure-it-outSkill
- fix-ciSkill
- fix-merge-conflictsSkill
- get-pr-commentsSkill
- graphSkill
- harborSkill
- hillclimbSkill
- howSkill
- hudSkill
- intentSkill
- interrogateSkill
- in-your-own-wordsSkill
- ketchupSkill
- ketchup-videoSkill
- launchSkill
- loftSkill
- maintain-formal-verification-skillSkill
- maintain-verification-skillSkill
- mapSkill
- minimal-code-disciplineSkill
- minimal-prose-disciplineSkill
- modelsSkill
- omc-doctorSkill
- omc-setupSkill
- personal-modeSkill
- omc-planSkill
- project-session-managerSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
7 rules. 0 are common, 5 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 14 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 47
Before writing logic: core types and data structures, scaffold-vs-feature sequencing, what concurrent actors share.
- Distinctive: line 59
When two or more attempted fixes share one assumption and all failed the same gate, stop writing fix three.
- Very specific: line 66
Before sequencing an addition, refactor, or rewrite, remove the dead weight first and build on the simpler base.
- Distinctive: line 105
When stateful logic branches a lot or repeats a shape assumption across files, encode the domain in a structure — a state machine, a typed …
- Distinctive: line 175
Before trusting, reporting, or acting on a measured number, find what limits it and rule out that it measured something other than the work …
- Distinctive: line 206
When delegating, prepend the delegate lane's standing principles to the brief.
- Distinctive: line 227
When this skill is active, name in the reply which principle shaped each non-obvious decision, in one clause.
“Common” is measured: kept by at least 6 in 10 of the 14 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Engineering Principles
Twenty-four principles distilled for OMC's lanes from the pstack principle library (credited above) and rewritten against OMC's own mechanisms. Each names when it applies and which lane carries it. This is a discipline skill: it shapes how work is done; it never gates delivery on its own.
When to use
- a fix has failed twice, or two fixes share one explanation
- debugging, verification, or benchmarking work is starting
- a design choice repeats a shape assumption across files
- delegation prompts are being written and a lane's standing principles should travel with the brief
When not to use
- a one-line mechanical edit with an obvious check
- a hotfix under time pressure — state which principle you skipped and why in one line, then proceed
Core
Laziness Protocol
Refactoring, sizing a diff, or tempted to add abstractions, layers, or
signal threading. Bias to deletion and the smallest change that solves the
problem. Composes with minimal-code-discipline. Carried by executor and
code-simplifier.
Foundational Thinking
Before writing logic: core types and data structures, scaffold-vs-feature
sequencing, what concurrent actors share. Carried by architect and
planner.
Redesign from First Principles
Integrating a new requirement into an existing design. Redesign as if the
requirement had been foundational from day one, instead of patching the old
shape around it. Carried by architect.
Attack the premise
When two or more attempted fixes share one assumption and all failed the
same gate, stop writing fix three. Write the shared assumption down, test
it directly, and let the result choose the next move. Carried by critic
during plan gap analysis and by debugger before a third retry.
Subtract before you add
Before sequencing an addition, refactor, or rewrite, remove the dead weight
first and build on the simpler base. Carried by executor and
code-simplifier.
Minimize reader load
Reviewing or shaping code that is hard to trace. Count layers and hidden
state, collapse one-caller wrappers, shrink mutable scope. Carried by
code-simplifier and writer.
Outcome-oriented execution
Planned rewrites and migrations with explicit phase boundaries. Converge on
the target architecture; do not preserve throwaway compatibility states.
Carried by planner.
Experience first
Product, UX, or feature-scope tradeoffs. Choose user delight over
implementation convenience, and say the tradeoff out loud when you take it.
Carried by designer.
Exhaust the design space
A novel interaction or architectural decision with no precedent. Build 2-3
competing options and compare before committing — the arena and
architect skills are this principle operationalized. Carried by
architect.
Build the lever
Any non-trivial repeated work. Build the tool that does or proves it —
codemod, script, generator — rather than doing it by hand a third time; the
tool is the artifact a reviewer reruns. Carried by executor.
Architecture
Model the domain
When stateful logic branches a lot or repeats a shape assumption across
files, encode the domain in a structure — a state machine, a typed model, a
registry, a reducer — instead of scattering conditionals. OMC's graph
runtime (graph skill) is this principle at full size. Carried by
architect.
Boundary discipline
Wiring validation, error handling, or framework adapters. Guards at system
boundaries, trust internal types, keep business logic pure. Carried by
architect and executor.
Type system discipline
Designing types or signatures in any typed language. Make illegal states
unrepresentable, brand primitives, parse external data at boundaries; in
TypeScript, typescript-best-practices grounds this in syntax. Carried by
architect and executor.
Make operations idempotent
Design commands, lifecycle steps, and loops that run amid crashes and
retries to converge to the same end state. OMC's team recovery replays
worker checkpoints idempotently; user code deserves the same property.
Carried by executor.
Migrate callers, then delete legacy APIs
Introducing a new internal API while old callers exist. Migrate every
caller and delete the old surface in one wave; a compatibility shim with no
expiry is a second system. Carried by executor.
Separate before serializing shared state
Concurrent actors might write the same file, branch, key, or object.
Eliminate the sharing first (a writer per worktree is OMC's house form);
locking is the fallback, not the first move. Carried by executor and
team.
Verification
Prove it works
Done means verified against the real artifact — the command's actual
output, the passing test, the rendered page — never a proxy and never "it
compiles". The verify skill owns the evidence format. Carried by
verifier.
Fix root causes
Trace each symptom to the first place the system went wrong and fix it
there. A guard for a state that should not exist is a workaround, not a
fix; it is legitimate only when that state itself is legitimate. Carried by
debugger; verifier treats symptom-level patches as incomplete work.
Sequence work into verifiable units
Multi-step work — sweeps, migrations, runs of similar edits — and how
commits and PRs stack. Small units that each end in a check, verified
before the next, ordered so the sequence proves itself. Carried by
planner and git-master.
Test behavior, not implementation
Assert what users of the code can observe. A test worth keeping fails when
the behavior breaks and passes when only internals change. Carried by
test-engineer.
Explain the number
Before trusting, reporting, or acting on a measured number, find what
limits it and rule out that it measured something other than the work you
think. Label every claim measured, inferred, or guess in the same sentence.
The benchmark-checklist skill owns the full procedure. Carried by
scientist and by any lane reporting performance.
Delegation
Guard the context window
Route bulk reading and large outputs to subagents; keep summaries in the lead thread. Context spent on a wall of log output is context not spent on judgment. Applies to every lane and to the orchestrator itself.
Never block on the human
Tempted to ask "should I do X?" on reversible work. Proceed, present the result, let the human course-correct; reserve questions for genuine product calls no experiment can settle. Carried by every lane.
Meta
Encode lessons in structure
The second time you write the same instruction, stop writing it a third
time: encode it as a lint rule, a runtime check, a script, or — in OMC — a
learned skill via skillify, or an enforcement via correct. Text rots;
structure reruns. Carried by skillify and self-improve.
Distribution map
When delegating, prepend the delegate lane's standing principles to the brief.
| Lane | Standing principles |
critic | Attack the premise |
debugger | Attack the premise, Fix root causes |
verifier | Prove it works, Fix root causes |
test-engineer | Test behavior, not implementation |
architect | Foundational Thinking, Redesign from First Principles, Model the domain, Exhaust the design space |
planner | Foundational Thinking, Outcome-oriented execution, Sequence work into verifiable units |
executor | Laziness Protocol, Subtract before you add, Make operations idempotent, Boundary discipline, Type system discipline, Migrate callers, Separate before serializing, Build the lever |
scientist | Explain the number |
designer | Experience first |
code-simplifier | Laziness Protocol, Subtract before you add, Minimize reader load |
writer | Minimize reader load |
git-master | Sequence work into verifiable units |
| every lane | Guard the context window, Never block on the human, Encode lessons in structure |
Reporting
When this skill is active, name in the reply which principle shaped each non-obvious decision, in one clause. When a principle and an explicit user instruction conflict, the user wins; say so in one sentence.
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.

