Context Creation (Brownfield Back-Fill)

What it does

Wraps an existing (brownfield) codebase in project state documentation, derived strictly from what is already on disk.

• Pre-flight: verifies .codearbiter/CONTEXT.md lacks the <!--INITIALIZED--> marker and that meaningful source exists; otherwise routes to decompose or to normal operation.
• Six parallel scouts: tech stack (A), infrastructure/CI (B), architecture (C), security posture (D), testing (E), data model (F). Each reads one slice and returns a structured report — file paths, line numbers, named values — never raw code.
• Confidence-graded synthesis: HIGH findings are written as fact, MEDIUM with an "inferred from…" note, LOW as [CONFIRM-NN] placeholders with an ID, the unknown, why it matters, and what resolves it.
• Gap interview: one targeted question per open item; vague answers are challenged, deferrals are logged in open-questions.md.
• Write and lock: emits CONTEXT.md, tech-stack.md, coding-standards.md, security-controls.md, open-questions.md, open-tasks.md, overrides.log (plus a conditional, marker-gated release-targets.md), then writes the initialization marker.

Who it's for

• Teams already running the codeArbiter plugin who need to onboard an undocumented legacy repo.
• Engineers who want a large codebase summarized without loading raw source into the orchestrator's context window.
• Anyone who needs a security-posture write-up that records paths and line numbers only, never secret values.

Caveat: it depends on the .codearbiter/ convention and plugin scripts such as {{PLUGIN_ROOT}}/hooks/taskwrite.py, so it is not practical as a standalone skill.

Examples

1. Onboarding a 5-year-old undocumented monorepo — run /create-context; the scouts read package.json, .github/workflows/, prisma/schema.prisma and produce drafts covering stack, build/test commands, and entity inventory.
2. Automatic startup routing — a session opens with CONTEXT.md present but no <!--INITIALIZED--> marker, so this skill takes over and finishes the half-done initialization.
3. Release target inference — with exactly one root manifest and one changelog, it writes a HIGH-confidence release-targets.md row; anything ambiguous becomes a [CONFIRM-NN] instead of a guess.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/arbiterForge/codeArbiter/HEAD/core/surface/skills/context-creation/SKILL.md

What I want: (describe your task here)

Install the skill I found on Claude Skill Mart.
Copy the core/surface/skills/context-creation folder from the GitHub repo arbiterForge/codeArbiter into my ~/.claude/skills/context-creation/.
When it's done, tell me in one line what this skill can do.