/type-mismatch

Types only exist until compile time. At runtime, any, casts, and untyped boundaries let anything through. This skill finds unvalidated runtime type assumptions.

What this is for

• JSON.parse() result used without validation (crash on unexpected shape)
• API response accessed without schema check (missing field = undefined)
• Type assertions/casts that hide real type mismatches
• Read-only skill. No code changes.

PROTECTION RULE - never ~/.claude/

Read-only skill. Guard required if write mode added later.

What You Must Do When Invoked

During analysis, assign a confidence level to each finding: proven (confirmed by evidence), likely (strong signal, needs review), or suspected (weak signal).

If /type-mismatch -help or /type-mismatch -h (without further arguments) is invoked: output the ## Usage section unchanged and stop.

Otherwise follow these steps in order, skipping none.

Step 1 - Help check

If invoked with -help or -h, output the ## Usage section unchanged and stop.

Step 2 - Clarify target

Clarify -ProjectDir. Get confirmation.

Step 3 - Scan

& "<SKILL_DIR>/scripts/type-assumption-scan.ps1" -ProjectDir "<path>"

Step 4 - Classification

Read each assumption:

• Crash: uncaught TypeError on null/undefined access
• Data-loss: wrong value used silently
• Safe: validated upstream or internal source

Step 5 - Write report

File type-mismatch-report.md in current working directory:

1. Summary - assumptions by risk level.
2. Assumption table - crash risk first. Per assumption: file, line, kind, origin, has validation, risk, recommendation.
3. Open questions.

Step 6 - Summarize

State report path, highlight assumptions that will crash in production.

Usage

/type-mismatch               # interactive
/type-mismatch <dir>         # scan project
/type-mismatch -help