cellm/skills/engineering/SKILL.md
CELLM engineering axioms plus pre-write discipline against silent assumptions, premature complexity, drive-by edits, and unverifiable goals. Activates when adding logic, changing control flow, or introducing dependencies in source files. Inactive on read-only operations, pure formatting/whitespace changes, comment-only edits, type-only refinements, and renames/moves with no semantic change.
npx skillsauth add murillodutt/cellm cellm/skills/engineeringInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Four non-negotiable principles for every line of code:
| Axiom | Meaning | Anti-Pattern |
|-------|---------|-------------|
| ONE VERSION | No v1/v2 coexistence. Migrate all, remove old. | if (version === 1), compatibility layers, convertOldToNew() |
| RESOLVE > SILENCE | Fix the cause, never suppress. | @ts-ignore, eslint-disable, silent catch, optimizeDeps.exclude |
| LAZY > EAGER | Load on-demand by default. | Eager imports of heavy deps, non-lazy components, blocking initialization |
| NOW > LATER | Can be done now? Do it now. | TODO comments for trivial fixes, deferred cleanup |
Bias: caution over speed. Trivial edits use judgment. For non-trivial work, pass these checks before the first edit:
| Guard | Required Before Editing | Blocks | |-------|-------------------------|--------| | THINK BEFORE CODING | Name assumptions, ambiguity, simplest safe interpretation. Push back when a simpler path exists. Stop and ask when confused. | Silent assumptions, confident wrong paths, hidden confusion | | SIMPLICITY FIRST | Add complexity only when needed now, with a real second consumer or explicit contract. If 200 LOC could be 50, rewrite. Senior-engineer test: would they call it overcomplicated? | Premature abstractions, speculative options, framework-shaped code for one case | | SURGICAL CHANGES | Every touched line traces to the request or to an orphan this change created. Match existing style. Mention unrelated dead code, do not delete it. | Drive-by refactors, style drift, adjacent rewrites, unrequested cleanup | | GOAL-DRIVEN EXECUTION | Define the falsifiable finish line before building. Bugs get reproduction; features get expected behavior; refactors get before/after green. State a step → verify plan for multi-step work. | "Make it work" loops, weak success criteria, closure without evidence |
Plan template for multi-step tasks:
1. [Step] -> verify: [check]
2. [Step] -> verify: [check]
3. [Step] -> verify: [check]
If a task is trivial, satisfy the guards silently and proceed. If the work is ambiguous, risky, broad, or architectural, state the guard outcome before editing.
defineAsyncComponent, dynamic imports, .client.vuePre-Write Discipline adapted from Tilly's behavioral guidelines (MIT, github.com/forrestchang/tilly-engineer-skills).
data-ai
Prose override — temporarily disable quantization and respond in readable prose. Use when relational density matters, for safety-critical explanations, onboarding handoffs, or when token economy is not the priority.
development
Govern explicit weekly Super PRs or maintainer-requested PR merges. Evaluates a 10-criterion readiness checklist and performs governed merge only when a user-requested PR is READY. Never creates or keeps permanent PRs. Use when: 'pr-check', 'pr-merge', 'merge this PR safely', 'is PR ready', 'guard merge', or /sk-git delegates pr-merge.
data-ai
Operational surface for the compress-llm Layer-1 token I/O compressor. Enable, disable, switch mode, and inspect status without editing config files. Use when tuning compression pressure for the current session or project.
tools
Generate structured upstream feedback for the CELLM engineering team. Produces evidence-first Markdown at docs/evidence/<date>-cellm-feedback-*.md for bugs, anti-patterns, deprecation gaps, and harness surprises, with optional atom registration via knowledge_ops. Use when: 'feedback for CELLM', 'send to CELLM team', 'register this as atom', 'document this anti-pattern', 'report this bug upstream'. Trigger proactively on MCP schema/runtime mismatches, mechanical edit loops (>=3 sequential edits), short deprecation windows (<6 weeks), or reusable harness surprises. Do NOT trigger for routine feature work or project-local bugs.