Task — engineering-spec@1

Release-readiness checklist for 1.0.0 at docs/release/

doneTASK-IMP-087
module improvement · class product · priority p1 · created 2026-07-16 · shipped 2026-07-16
depends on none · blocks none

TASK-IMP-087: Release-readiness checklist for 1.0.0 at docs/release/

Summary

VERSION says 1.0.0 but nothing defines what "ready to tag" means. Write the release checklist as a living tracked document: every line carries an owner, a state (open, checked, or waived with reason), and the verification command or evidence link where one exists. Seed it with the seven IMP-15 lines, the three operator decisions recorded today (IMP-06/07/11, now pre-release implementation items), and the channel-freshness matrix from the 2026-07-16 research.

Problem

The pre-1.0.0 hardening run closed real payload gaps found only by running the workflows against a live consumer repo. The remaining risk is release mechanics that have never been exercised: the npm package has never been packed, the plugin zip never loaded into a live session end-to-end from a release asset, the tag flow never dry-run. A checklist nobody wrote is a release gate nobody holds.

Proposed Solution

docs/release/RELEASE-CHECKLIST.md: a table (line, owner, state, evidence) grouped into (a) code readiness - S1 handoff items landed or waived with reason; (b) artifact readiness - npm pack dry-run of @cyberskill/cyberos plus an npx smoke on a scratch repo, plugin zip loaded into a live Claude Code or Cowork session with the three commands triggered once, release.yml tag-flow dry-run attaching payload assets; (c) channel readiness - the agent-surface matrix re-verified against current tool conventions (including .devin/rules/ preference and the shared .agents/skills/ dir); (d) docs readiness - CHANGELOG release section, GUIDE pass, fresh-clone consumer test (git clone sachviet && npm ci && npm run coverage green); (e) the three decided items IMP-06/07/11 implemented or explicitly deferred past 1.0.0. English, no secrets, cross-linked to IMPROVEMENT_HANDOFF.md and the batch evidence.

Alternatives Considered

Success Metrics

Scope

In scope: the checklist document, its cross-links, the recorded presence checks.

Out of scope / Non-Goals

Dependencies

AI Authorship Disclosure

1. Description (normative)

2. Acceptance criteria

3. Edge cases

4. Out of scope / non-goals

Duplicated intentionally with ## Scope for template conformance: executing lines, implementing the decided items, CI automation.

5. Protected invariants this task must not weaken

End of TASK-IMP-087.

Audit

TASK-IMP-087-release-checklist spec audit - audit_rubric@2.0 (machine floor + judgment)

Machine floor: task-lint.mjs (first governed use). Judgment families walked by the model: QA semantics (metrics grounded in run evidence with baseline/target/deadline; alternatives distinct; scope subsections; no unsourced numerics; no cross-team claims), SAFE content (sourced blocks, no injection markers), TRACE semantic sufficiency (cited tests/ops evidence genuinely prove their clauses), COND-004 content (three labeled bullets truthful).

ISSUE ISS-001 (TRACE-002, wontfix-info): ACs are ops-verified with recorded-grep rationale (a single operator markdown document; suite out of scope by design). Sufficient. ISSUE ISS-002 (QA-004, wontfix-info): the primary metric measures document shape, not release outcome - correct on purpose: working the lines IS the release, operator-owned; stated in Success Metrics.

SUMMARY verdict: pass issues_open: 0 issues_human: 0 iterations: 2 next_action: ship

§gate-log

Populated during implementation (ship-tasks testing phase).