Skip to main content
skill CX Codex

grounding-before-coding — Claude Code skill: inspect code and data

The grounding-before-coding skill makes the agent inspect code and data before changing a project or explaining a bug. It fits unfamiliar repositories and non-trivial work.

no license 25,600 installs ★ 6 updated 2026-09-11

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add riekelt/principal-engineer --skill grounding-before-coding -a claude-code
Open on GitHub

What grounding-before-coding does

grounding-before-coding sets a research step before implementation. The agent traces actual entry points, reads the code behind names, checks the relevant data, and supports important claims with file:line, query results, or command output. For a bug, it must observe the failure before changing the code and inspect every route into a shared area.

The skill also records invariants, existing guards, and repository conventions that a change must preserve. It covers investigations where no edit is planned yet. The required background is the principal-engineering skill, so that dependency belongs in the setup plan.

What it helps with

  • understand an unfamiliar repository
  • reproduce a bug before editing code
  • trace every caller of shared logic
  • support conclusions with code evidence
  • record invariants before a change

Example prompts for the agent

  • Before changing anything, inspect the project tree and support each key finding with a file:line reference.
  • Investigate why an export is sometimes empty: reproduce the failure first, then trace the code and data.
  • Find every route into this shared code and list the invariants the fix must preserve.

How to install grounding-before-coding

1Skills CLI (Vercel)

npx skills add riekelt/principal-engineer --skill grounding-before-coding -a claude-code

2Manually: put the files in place

~/.claude/skills/grounding-before-coding/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/grounding-before-coding/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about grounding-before-coding

How do I install grounding-before-coding in Claude Code?

Run npx skills add riekelt/principal-engineer --skill grounding-before-coding -a claude-code, then check ~/.claude/skills/grounding-before-coding/SKILL.md.

What does grounding-before-coding do?

It makes the agent inspect code and data before a change or conclusion, support key claims with evidence, and reproduce failures before fixes.

What does grounding-before-coding require?

The skill requires principal-engineering as background.

Similar skills

All skills for Agent setup: memory, context, orchestration →

From the same set11

The whole set →
writing-unit-tests writing-unit-tests guides an agent through writing or revising unit tests around observable behavior. Useful for flaky tests, confusing cases, and new test coverage. CX Codex +1 25,700installs keeping-one-source-of-truth keeping-one-source-of-truth — an agent skill that keeps each fact in exactly one place and stops copies from drifting apart. Useful the moment a value begs to be pasted elsewhere. CX Codex +1 25,600installs testing-changes testing-changes defines test requirements for code changes, enforcing regression tests for bug fixes, edge-case coverage, and strict checks against empty test diffs. CX Codex +1 25,600installs guarding-architecture guarding-architecture enforces system boundaries by turning structural invariants into breaking build tests. Useful when code crosses module limits or adds unwanted cross-package imports. CX Codex +1 25,600installs verifying-before-done verifying-before-done — an agent skill that forces a check on the changed code's real surface and a pasted verify output before any completion claim goes out. CX Codex +1 25,600installs adding-dependencies The adding-dependencies skill helps decide whether a project needs a dependency and how to assess and update it. Use it before adding a package, SDK, or GitHub Action. CX Codex +1 25,500installs handling-failures handling-failures — a skill that forces AI assistants to write honest error paths. Every failure either gets logged and rethrown, returns a typed error, or shifts the system into an explicit degraded mode. CX Codex +1 25,500installs operating-safely operating-safely sets safety guards for destructive commands, secret handling, and parallel edits. It activates before dangerous operations like deleting data or restarting live processes. CX Codex +1 25,500installs

Project sponsors

Companies that keep this analytics open