CLI Audit Shell

Optimization: This skill uses on-demand loading. Heavy content lives in references/ and is loaded only when needed.

Language rule: Skill instructions are written in English. This skill targets Bash scripts exclusively. When generating user-facing output, detect the project's primary language (from README, comments, docs, commit messages) and produce the report in that language. If the project is bilingual, ask the user which language to use before proceeding.

Audit Shell — Shell Quality Index (SQI)

"Shell should only be used for small utilities or simple wrapper scripts." — Google Shell Style Guide

Core Principles

1. Evidence-based — every finding needs a file:line reference. No vague "the script could be better"
2. Proportional — a 500-line script doing LVM partitioning matters more than a 10-line wrapper. Scale to script complexity
3. Beyond shellcheck — shellcheck catches syntax; this skill catches semantic anti-patterns (dead code under set -e, redundant package checks, injections in heredocs)
4. Named anti-patterns — use the anti-pattern names from references/anti-patterns.md in findings
5. Positive reinforcement — always highlight good practices found
6. Complexity-adapted tooling — simple problems deserve simple tools (sed), complex problems deserve proper tools (kustomize). Read references/tooling-ladder.md
7. Gotchas — read ../gotchas.md before producing output to avoid known mistakes

Input

$ARGUMENTS is the target to audit (file path, directory, or empty for auto-detect).

• If a specific .sh file: audit that file deeply
• If a directory: audit all .sh files and shebanged scripts in it
• If empty: find all shell scripts in the project (glob **/*.sh + grep #!/bin/bash or #!/usr/bin/env bash)

12-Dimension Framework

Score each dimension 0.0-1.0, then compute a weighted SQI. Read references/categories.md for detailed checklists per category.

Workflow

Step 1 — Discover, sample, and detect language

Glob for shell scripts: **/*.sh, **/Makefile, shebanged files. For broad audit, prioritize: entry points (setup.sh, install.sh, deploy.sh), longest scripts, most-changed scripts (git log).

Detect output language: git log --oneline -10 + head -20 README.md. Report in the project's language (French commits → French report).

Step 2 — Detect script context

Classify each script:

Step 3 — Run shellcheck first

shellcheck -x -S style <file>

If shellcheck is not installed, note it and proceed with manual analysis. Shellcheck findings go into the report but don't duplicate them in the dimension scores — this skill adds what shellcheck cannot see.

Step 4 — Score all 12 dimensions

Read references/categories.md for detailed checks. For each category: collect evidence, assign score 0.0-1.0, note specific file:line findings.

Step 5 — Detect anti-patterns

Read references/anti-patterns.md for named anti-patterns with detection heuristics. These are the semantic patterns shellcheck cannot find.

Step 6 — Assess tooling fit

Read references/tooling-ladder.md for the complexity-adapted tooling decision tree. Flag scripts that use sed/awk for YAML when they should use yq/kustomize. Flag scripts that use kustomize for a 3-variable substitution when sed would suffice.

Step 7 — Compute SQI and generate report

Read references/scoring.md for the SQI formula, severity classification, and benchmarks.

SQI = Sigma(wi x si) / Sigma(wi) x 10

Finding tier and confidence semantics are canonical in ../shared/triage.md (Tier 3/2/1 + GRADE + triangulation). S-dimension findings carry tier and confidence so cli-cycle can aggregate without re-parsing.

Output Format

# Shell Audit -- {project-name}

**Target**: [file/directory] | **Scripts found**: [N] | **Date**: [date]
**SQI Score**: X.X/10 -- {verdict}
**Shellcheck**: [N warnings, M errors] (or "not available")

## Scores by Category

| # | Category | Weight | Score | Weighted | Key findings |
|---|----------|--------|-------|----------|-------------|
| S1-S12 rows with 0.0-1.0 scores... |
| | **SQI** | **100%** | | **X.X/10** | |

## Anti-Patterns Detected
| # | Pattern | Severity | File:Line | Recommendation |

## Critical Violations (must fix)
### [Category]: [violation title]
- **File**: `path/to/file:123`
- **What**: [description]
- **Why**: [named principle it violates + Google Shell Style Guide reference]
- **Fix**: [concrete code suggestion]

## Flags (should fix)
[same format]

## Tooling Recommendations
| Current tool | Used for | Recommended | Why |
[complexity-adapted tooling analysis]

## Good Practices Found
[positive reinforcement]

## Recommended Next Steps
1. [highest-impact fix first]
2. [second]
3. [third]

What this skill does NOT do

• Does not replace shellcheck — it complements it with semantic analysis shellcheck can't do
• Does not fix scripts — it reports. Use the findings to guide refactoring
• Does not audit non-shell code — YAML, Dockerfiles, Terraform go to cli-forge-infra
• Does not audit test strategy — use cli-audit-test for that
• Does not audit code topology — use cli-audit-tangle for call graph analysis

Integration with other cli-* skills

Dynamic Handoffs

Rule: Recommend, don't auto-execute.

Reference Sources

• Google — Shell Style Guide (primary authority)
• Wooledge — BashGuide, BashFAQ, BashPitfalls
• ShellCheck — SC1000-SC9999 rule database
• Kfir Lavi — Defensive Bash Programming
• Dylan Araps — Pure Bash Bible
• Greg Wooledge — Bash Idioms (O'Reilly)
• SixArm — Unix Shell Script Tactics (XDG compliance, NO_COLOR standard, portability catalog)
• clig.dev — Command Line Interface Guidelines (CLI ergonomics, --json, --dry-run)
• no-color.org — NO_COLOR standard (color output gating)
• POSIX.1-2024 — Shell Command Language specification