find-edgecase-slop

Find Edge-Case Slop

Look for places where we have hard coded or spiked the complexity for an edge case. Report your findings.

This skill is read-only planning: enter plan mode, scan the CWD repo, and materialize every hit as a todo. Do not edit code or start cleanup unless the user explicitly asks after reviewing the plan.

Enter plan mode (first action)

1. If the session is not already in plan mode, switch into it (Cursor: SwitchMode → plan; other harnesses: the equivalent read-only / planning posture if available). Explain briefly: edge-case slop findings must land as a plan with exhaustive todos before any cleanup.
2. If already in plan mode, stay there — do not bounce modes.
3. Stay in plan mode through the scan and todo materialization. Only leave plan mode if the user later asks to clean up (then switch to agent / implementation mode).

Scope

• Repo = CWD. Resolve the git repo rooted at (or containing) the current working directory. Scan that repo only — not sibling repos under ~/code.
• If CWD is not inside a git repo, stop and say so. Do not invent a target.
• No path argument required. Optional focus hints from the user (a path, module, or subsystem) narrow the scan; otherwise cover the whole repo.

What counts as edge-case slop

Flag hard-coded or spiked complexity for a rare / one-off path — not every branch, guard, or validation. Prefer hits that look like debt, not legitimate domain rules.

Smell patterns (non-exhaustive):

• Special-cased branches / if ladders for a single weird input, customer, ID, locale, browser, or date
• Magic literals, allowlists, denylists, or override maps whose only job is one rare case
• Near-duplicate copies of a happy-path function with one extra twist
• Comment-marked spikes (// edge case, // hack for …, // special case for X, FIXME / HACK tied to a rare path)
• Compatibility shims or one-off adapters that exist only for an obsolete or singular caller
• Test-only fixtures that force production code to grow a permanent special case

Do not flag ordinary validation, null checks, standard error handling, or complexity that clearly encodes a durable product rule — unless it's clearly a one-off spike that should have been a general mechanism.

Procedure

1. Enter plan mode (above).
2. Confirm the CWD git root (git rev-parse --show-toplevel).
3. Scan the repo (respecting any user focus hint). Use search + targeted reads; prefer evidence (file:line, short quote of the spike) over vibes.
4. Report findings briefly in the plan (path + why it's edge-case slop).
5. Immediately materialize exhaustive todos (below). The plan is incomplete until every finding is a todo.
6. Pause. Do not edit or clean up unless the user asks.

Materialize the full todo list (still plan mode)

After the scan, create a todo for every finding via the harness todo tool (TodoWrite in Cursor; equivalent elsewhere). Mandatory — not optional polish.

• One todo per finding, not one vague "clean up edge cases" bucket.
• Content must stand alone — path (ideally file:line) + short why ("Hard-coded override for customer X in foo.ts:42"). Vague todos fail this gate.
• Exhaustive — every reported finding gets a todo; no silent omissions. If the scan finds nothing, say so in one sentence and leave the todo list empty (do not invent filler todos).
• Mark all todos pending. Only flip them when the user later asks to act and work lands.

Don't

• Don't edit code, open cleanup PRs, or leave plan mode unless the user asks.
• Don't expand into a general complexity / refactor census — stay on edge-case spikes.
• Don't invent findings to fill the list; empty is a valid result.
• Don't scan outside the CWD git repo.