Use when the user wants to know what will break if they change something, or needs safety analysis before editing code. Examples: "Is it safe to change X?", "What depends on this?", "What will break?"
GitNexus Impact Analysis is a code-safety skill that answers the question "what will break if I change this?" before an edit or commit. It leans on the GitNexus dependency index (via an MCP `impact` tool or a `node .gitnexus/run.cjs` / `npx gitnexus` CLI fallback) to walk the caller graph of a target symbol and report its blast radius. The problem it solves is the uncertainty around modifying shared code: rather than guessing who depends on a function, the skill surfaces direct callers, indirect dependents, and affected execution flows so changes can be made deliberately.
The workflow runs `impact` upstream on a target, reads the repo's `processes` resource to see which execution flows are touched, then runs `detect_changes` for a git-diff-based pre-commit check. Output is graded by depth: d=1 items "will break," d=2 are "likely affected," d=3 "may need testing," with confidence scores and a risk scale from LOW to CRITICAL. The skill is notably careful about epistemic honesty: it flags an UNKNOWN state when the walk cannot resolve callers (dynamic dispatch, plain-object access, cross-language calls) and warns against treating zero callers as "safe to delete" without a text-search confirmation.
It targets developers and AI coding agents making non-trivial changes to unfamiliar or large codebases, especially before commits or refactors. Typical uses are assessing whether a function is safe to modify, mapping who uses a piece of code, and generating a risk report. It is a single-file SKILL.md that documents how to drive the external GitNexus tool rather than shipping the analysis engine itself.
It analyzes the dependency blast radius of a code symbol, listing direct callers (which will break), indirect dependents, and affected execution flows, then assigns a risk level so you can decide whether a change is safe.
It uses the GitNexus MCP `impact` and `detect_changes` tools, or a CLI fallback via `node .gitnexus/run.cjs` (or `npx gitnexus`). The repository must be indexed; if the index is stale you run `analyze` first.
Results are graded by depth: d=1 will break, d=2 is likely affected, d=3 may need testing, each with a confidence score. An affected-symbol count maps to LOW/MEDIUM/HIGH/CRITICAL risk.
No. The skill explicitly marks zero callers as UNKNOWN rather than LOW, because the index may miss dynamic dispatch or cross-language calls. It advises confirming with a text search before treating a symbol as unused.
No. It is a SKILL.md that documents how to drive the external GitNexus tool and interpret its results; the indexing and graph queries are performed by GitNexus itself.
Quick Setup:
.claude/skills/Repository
abhigyanpatwari/gitnexus