Skip to content
Skills Library
Programming · Review & audit

dead-code-audit

/dead-code-audit

Use when auditing Julie for dead code, test-only code paths, stale helpers, graph gaps, or cleanup candidates before deleting, privatizing, or simplifying code.

Repository
julie ↗
Licence
MIT
Last repository change
02.10.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
2
Scripts
1
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 02.10.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What to check before installing
  • 1 scripts in this package. Review them before installing, as Claude may execute them.

Scripts: scripts/dead_code_inventory.py

What it does and when to use it

Use when auditing Julie for dead code, test-only code paths, stale helpers, graph gaps, or cleanup candidates before deleting, privatizing, or simplifying code.

Original author description

Use when auditing Julie for dead code, test-only code paths, stale helpers, graph gaps, or cleanup candidates before deleting, privatizing, or simplifying code.

How it works

  1. Scope
  2. Inventory
  3. Triage Labels
  4. Required Verification
  5. Reading Inventory Sections
  6. False-Positive Traps
  7. Deletion Rules
  8. Report Format
Original SKILL.md headings · 148 lines, 779 words
SKILL.md preview
---
name: dead-code-audit
description: Use when auditing Julie for dead code, test-only code paths, stale helpers, graph gaps, or cleanup candidates before deleting, privatizing, or simplifying code.
allowed-tools: mcp__julie__manage_workspace, mcp__julie__fast_search, mcp__julie__fast_refs, mcp__julie__deep_dive, mcp__julie__blast_radius, mcp__julie__get_context, Bash
---

# Dead Code Audit

Use Julie's index as evidence, then verify with Julie tools before any edit. A zero-reference row is a lead, not a verdict.

## Scope

This skill is for Julie development:
- dead code and stale helper audits
- code referenced only by tests
- public APIs with no product callers
- graph linkage gaps that look like dead code
- cleanup campaigns before deletion, privatization, or module splitting

Do not use this as a blind deletion workflow. Dynamic dispatch, trait impls, CLI entry points, MCP handlers, generated fixtures, parser registries, macro usage, and cross-language extraction paths all create false positives.

## Inventory

1. Get the current workspace id:

```text
manage_workspace(operation="status")
```

2. Run the bundled inventory script from the repo root:

```bash
python3 .claude/skills/dead-code-audit/scripts/dead_code_inventory.py --workspace-id <workspace_id> --limit 40
```

Use `--db <path>` if the workspace's index is not under the default `~/.julie/indexes/<workspace_id>/` location (e.g. a standalone `.julie/` project index). Use `--json` when another script will postprocess the result. The default report excludes fixtures, docs, examples, and data/config languages. Add `--include-fixtures` or `--include-data-languages` when that scope matters.

3. Also run early-warning signals:

```bash
./target/debug/julie-server signals --workspace . --fresh --limit 40
```

Build debug first if needed. Keep this read-only.

## Triage Labels

Assign one label to each candidate:

| Label | Meaning |
| --- | --- |
| `delete` | No product reachability, no external dispatch role, no useful test-only value |
| `make-private` | Used in product code, but exported wider than needed |
| `merge-into-caller` | Single product caller and no independent concept |
| `test-fossil` | Product symbol exists only to satisfy tests |
| `graph-gap` | Julie failed to link real product usage |
| `keep` | Valid entry point, trait hook, registry item, dynamic target, or public surface |
| `needs-design-review` | Risky or ambiguous enough to discuss before edits |

## Required Verification

For every candidate before recommending deletion:

0. Confirm the index is fresh:

```text
manage_workspace(operation="health", workspace_id="<workspace_id>", detailed=true)
```

If files changed since indexing, refresh or rebuild the debug index before trusting the report.

1. Resolve the exact symbol:

```text
fast_search(query="<name>", file_pattern="<narrow path>", workspace="<workspace_id>")
```

2. Check references with identifier fallback:

```text
fast_refs(symbol="<name>", include_definition=true, limit=200, workspace="<workspace_id>")
```

If the name is overloaded, disambiguate with the file path and rerun.

3. Inspect the symbol and its role:

```text
deep_dive(symbol="<name>", context_file="<path>", depth="context", workspace="<workspace_id>")
```

Look for trait implementations, `#[cfg(test)]`, CLI/MCP registration, parser factory registration, annotation-driven entry points, and cross-language extractor contracts.

4. Search for non-graph usage:

```text
fast_search(query="<symbol-name>", limit=50, workspace="<workspace_id>")
```

Check strings, config, route names, command names, JSON/TOML/YAML keys, shell scripts, CI files, macro sites, and plugin manifests.

5. Check impact before editing:

```text
blast_radius(file_paths=["<path>"], include_tests=true, max_depth=2, workspace="<workspace_id>")
```

If the candidate lives in shared indexing, database, search ranking, workspace registry, parser extraction, or MCP handling code, escalate to design review before deletion.

## Reading Inventory Sections

- **Test-Only Relationship Refs**: strongest fossil signal. Tests call it, product code does not. Still verify trait hooks and registries.
- **Zero Relationship Refs And No Production Identifier Hits**: deletion candidate pool. High false-positive risk for dynamic entry points.
- **Likely Graph Gaps**: not dead-code candidates. These are cases where raw identifiers exist in product code but relationship edges are missing. Common names such as `new`, `run`, `handle`, `setup`, and `to_string` are noisy here.
- **Cfg-Test Markers In Source Paths**: test helpers embedded in source files. Classify as test cleanup, not product dead code.

## False-Positive Traps

- `reference_score = 0` means no scored incoming signal, not unused.
- `Contains` relationships are structure, not usage.
- Name-only identifier hits are leads, not proof.
- Wildcard imports can hide test-only calls from graph output. Use targeted content search before calling a symbol definition-only.
- Metadata is language-specific. Interpret keys such as `test_role`, `is_test`, decorators, impl metadata, and test linkage with context.
- Public symbols can be entry points through MCP registration, CLI dispatch, web routes, framework callbacks, parser factories, trait implementations, DI, or runtime naming conventions.

## Deletion Rules

Before editing code:
- write or identify the regression test that protects live behavior
- remove tests that only preserve the fossil behavior
- run the narrowest relevant test first
- after a batch, the lead session runs `cargo xtask test dev`

Never add a new test whose only purpose is to keep a questionable symbol alive.

## Report Format

```text
Dead Code Audit: <scope>

Candidate: <symbol> at <path>:<line>
Evidence:
- inventory section: <section>
- fast_refs: <prod refs>, <test refs>
- deep_dive role: <entry point / helper / trait hook / unknown>
- blast_radius: <summary>

Decision: <label>
Action: <delete / make private / merge / keep / review>
Verification: <test command>
Confidence: <1-100>
```
Files in this package (2)
  • SKILL.md 6 KB
  • scripts/dead_code_inventory.py 14 KB

Versions

RepositoryScoreLicence
This page · anortham/julie100MIT
tansuasici/claude-code-kit100MIT

Database snapshot: 13.11.2026 · package version ed7f68944d92 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100editing
    Use before making changes to existing files. Routes to Julie's edit_file tool without a Read + Edit loop. Trigger on fix, change, update, modify, refactor, rename, replace, add, remove, move, or any task involving an existing file.
  • 90audit
    Use when the user explicitly asks for a structural audit, architecture review, dead-code report, test-gap analysis, or evolution hotspot scan. Heavier and noisier than the other raysense skills — only run on demand, not as part of the routine edit loop.
  • 100find-dead-code
    Find unused functions, classes, imports, and unreachable code paths. Use when the user asks to find dead code, unused code, cleanup candidates, or wants to reduce codebase size.
  • 100exhaustive-systems-analysis
    Perform evidence-driven, multi-subsystem audits of real codebases to find correctness bugs, race conditions, security gaps, stale documentation, dead code, and production-readiness risks. Use when asked to audit a system end-to-end, verify agent-written code before shipping,…
  • 100code-verification
    Use after every code change, after writing tests, or before marking a task complete, and when the user says verify, quality gate, or audit code. Runs a 7-category mechanical check for TDZ, import, dead code, and mock leakage bugs.
  • 100configure-dead-code
    Dead-code detection: Knip, Vulture, cargo-machete, deadcode. Use when setting up unused-code scanning, migrating tools, or adding dead-code CI checks.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.