coverage-gate-author

name	coverage-gate-author
description	Test coverage gate (testing → done) — run the project's test suite + measure coverage on the files touched by the current task (per the `git diff` since the task's `implementing` status was set). Emits a coverage-gate@1 artefact: raw terminal output of the coverage tool, per-file coverage %, list of files below 90 %, list of edge-case-matrix rows without a corresponding test. Used by `chief-technology-officer/ship-tasks` during the `testing` phase to gate the `testing → done` transition (per `modules/skill/contracts/task/STATUS-REFERENCE.md` §1.1). Use when user asks to "draft a coverage gate" or "create the coverage gate". Do NOT use for "audit existing coverage gate" (use coverage-gate-audit instead). Do NOT use for spec correctness — that is `task-audit`'s job, run during the `draft → ready_to_implement` transition; the two gates are deliberately separated so spec correctness can be verified before any implementation work begins.
license	Apache-2.0
metadata	{"version":"1.0.0","module":"skill","stage":"e","cyberos-template":"coverage-gate@1","cyberos-rubric-target":"coverage_gate_rubric@1.0"}
allowed_memory_scopes	{"read":["project:*","module:*"],"write":["project:task/{task_id}/coverage-gate"]}
audit	{"row_kind":"coverage_gate_authored","required_fields":["task_id","files_touched","files_below_90pct","total_tests_run","tests_failed","ecm_rows_uncovered"]}
inputs	[{"name":"task","format":"task@1","required":true},{"name":"edge_case_matrix","format":"edge-case-matrix@1","required":true}]
outputs	[{"name":"report","format":"coverage-gate@1"}]
triggers	["workflow `chief-technology-officer/ship-tasks` testing phase (step 23 — first author call; step 24 audit)"]
blockers	["no coverage tool configured in repo — must be resolved first","test framework is broken — diagnose before running this skill"]
untrusted_inputs	{"wrap_in_marker":"untrusted_content","injection_scan":"required","on_marker_hit":"surface_to_human"}

coverage-gate-author

1. What it does

1. Reads the task's building status timestamp from BACKLOG.md.
2. git diff --name-only <building_ts>..HEAD → the touched-files set.
3. Picks the right coverage tool per language (rust: cargo tarpaulin or cargo llvm-cov; python: pytest --cov; node: vitest --coverage).
4. Runs the full test suite.
5. Reports per-file coverage for every file in the touched set.
6. Cross-references the edge-case matrix: every row's planned_test must exist + must have passed.
7. type: bug only — runs the regression proof (§4). Reads type from the task's frontmatter; skips entirely for any other type.

2. Output schema

# coverage-gate@1
task_id: task-<MODULE>-<NNN>
generated_at: <ISO-8601>
language: rust | python | typescript | mixed
tool: tarpaulin | llvm-cov | pytest-cov | vitest | ...
total_tests_run: <int>
tests_failed: <int>
overall_coverage_pct: <float>
files_touched:
  - { path: "...", coverage_pct: 92.3, lines_covered: ,   }
 []
 [, , ]