sarus

name	sarus
description	Repo-wide scan to find question patterns that commonly require interpretation before answering.
output	{"format":"prose"}

Repo Unclarity Scan

Input: $ARGUMENTS

---

Overview

Repo-wide scan to find question patterns that commonly require interpretation before answering. This is the broad survey that identifies where to focus detailed rewrite work.

Steps

Step 1: Scan for Unclarity Patterns

Search across all target files for these patterns:

Pattern	Regex/Signal	Why Problematic
Parenthetical additions	(...) in questions	Bundles extra checks invisibly
If-yes/if-no branching	"if yes", "if no"	Hides conditional logic in text
Compound "or" questions	"X or Y?"	Unclear if asking for choice or checking either
Multi-sentence questions	Two+ sentences ending in ?	Usually bundled questions
Undefined term hotspots	Same noun used 5+ times without definition	Everyone reads it differently
"If applicable"	Phrases like "where relevant"	Who decides applicability?
Subjective adjectives	"appropriate", "adequate", "good"	No shared standard
Implicit quantifiers	"The requirements", "all steps"	Which ones specifically?

Step 2: Catalog Findings

For each finding:

FINDING:
File: [path]
Line: [number]
Pattern: [which pattern matched]
Text: [the problematic text]
Severity: [H/M/L — based on how central the file is]

Step 3: Prioritize by Centrality and Impact

Not all findings are equally important. Prioritize:

Priority Tier	Criterion
P1 — Critical	Core gates (affect every project)
P2 — High	Procedures referenced by gates
P3 — Medium	Frequently used procedures
P4 — Low	Rarely used or peripheral procedures

Step 4: Group by Pattern

Aggregate findings to see which patterns are most prevalent:

Pattern	Count	Files Affected	Priority Distribution
[pattern]	[N]	[N files]	P1: [N], P2: [N], P3: [N]

This reveals systemic issues vs one-off problems.

Step 5: Generate Rewrite Queue

Order fixes by impact:

1. P1 findings with most common patterns first (highest impact per fix)
2. P2 findings next
3. Batch similar patterns together (fix all parenthetical bundles at once)

REWRITE QUEUE:
| # | File | Line | Pattern | Priority | Estimated Fix |
|---|------|------|---------|----------|--------------|
| 1 | [path] | [line] | [pattern] | P1 | [brief fix description] |
| 2 | ... | | | | |

Step 6: Report

REPO UNCLARITY SCAN:
Files scanned: [N]
Lines scanned: [N]
Findings: [N total]

By pattern:
| Pattern | Count | Most Affected Files |
|---------|-------|-------------------|
| [pattern] | [N] | [top 3 files] |

By priority:
P1 (critical): [N]
P2 (high): [N]
P3 (medium): [N]
P4 (low): [N]

Systemic issues: [patterns appearing in 10+ files]
Top rewrite targets: [files with most P1/P2 findings]

Rewrite queue: [N items, starting with highest impact]

When to Use

• When doing large-scale quality improvements
• When standardizing question style across gates/procedures
• When searching for the highest-leverage rewrite targets
• → INVOKE: /saqrc (question rewrite chains) for fixing found issues
• → INVOKE: /saadag (detectors and generators) for detailed analysis of specific files

Verification

• All target files scanned
• Multiple pattern types checked (not just one)
• Findings prioritized by centrality
• Patterns aggregated to identify systemic issues
• Rewrite queue ordered by impact