| name | edge-case-matrix-author |
| description | Generate a structured edge-case-matrix@1 for a task before implementation. Enumerates: null/empty inputs, extreme bounds (off-by-one, integer overflow, time-zone DST, leap second, Unicode normalisation), malformed payloads (truncated, oversized, non-UTF-8, type-confused), concurrent race conditions (double-submit, double-acknowledge, cross-tenant cross-talk, RLS escape), security-class entries (auth bypass, injection, token replay), and degradation modes (downstream slow, downstream unreachable, partial write). One matrix row per category-and-trigger with a pointer to the test that will cover it. Used by chief-technology-officer/ship-tasks as step 5. Use when user asks to "draft a edge case matrix" or "create the edge case matrix". Do NOT use for "audit existing edge case matrix" (use edge-case-matrix-audit instead). |
| license | Apache-2.0 |
| metadata | {"version":"1.0.0","module":"skill","stage":"e","cyberos-template":"edge-case-matrix@1","cyberos-rubric-target":"edge_case_matrix_rubric@1.0"} |
| allowed_memory_scopes | {"read":["project:*","module:*"],"write":["project:task/{task_id}/edge-case-matrix"]} |
| audit | {"row_kind":"edge_case_matrix_authored","required_fields":["task_id","total_rows","categories_covered","security_class_count","planned_test_paths"]} |
| inputs | [{"name":"task","format":"task@1","required":true},{"name":"context_map","format":"repo-context-map@1","required":false}] |
| outputs | [{"name":"matrix","format":"edge-case-matrix@1"}] |
| triggers | ["any task moving from `ready_to_implement` → `implementing`","workflow `chief-technology-officer/ship-tasks` step 5"] |
| blockers | ["task acceptance criteria are ambiguous — escalate to chief-product-officer","no test framework declared in repo — must be resolved first"] |
| untrusted_inputs | {"wrap_in_marker":"untrusted_content","injection_scan":"required","on_marker_hit":"surface_to_human"} |
edge-case-matrix-author
1. Purpose
Produce a single artefact — the edge-case-matrix@1 — that captures every boundary condition the implementation must address, before any code is written. The matrix is the contract between the task's acceptance criteria and the test suite; the coverage-gate skill (which runs after implementation) reads the matrix to verify every row has a corresponding test.
2. Output schema
task_id: task-<MODULE>-<NNN>
generated_at: <ISO-8601>
total_rows: <int>
rows:
- id: ECM-001
category: NULL_INPUT | BOUNDARY | MALFORMED | CONCURRENT | SECURITY | DEGRADATION
trigger: "<one-sentence description of what produces the edge case>"
expected: "<one-sentence description of correct behaviour>"
severity: critical | high | medium | low
planned_test: "<absolute path or test name where this will be covered>"
3. Quality gates
- Every category has ≥1 row (the audit-companion will fail if any is empty).
- SECURITY rows have a
planned_test pointing at a real test file (not TBD).
- DEGRADATION rows describe both detection and recovery.
total_rows ≥ 8 for any task rated priority.
