19c1f54a2a
* test(#2316): fail-first tests for ghost REQ-IDs, v-heading over-match, all-orphan gap check
Red phase: 4 of 10 fail against current source (#2316-1 ghost-ID warning,
-3 requirements_updated honesty, -4a v1-heading suppression, -6b all-orphan gap
rows). The other 6 are controls/boundaries that must keep passing — including the
#1159 deferred-heading guard and the literal "TBD" placeholder boundary.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SLufH5sDuqA1AiEGu45cuA
* fix(#2316): stop phase complete silently dropping ghost requirement IDs
phase complete parses a phase's `**Requirements**:` line from ROADMAP.md and
reconciles it into REQUIREMENTS.md. When a cited ID was registered nowhere, every
branch degraded to a no-op and the report was indistinguishable from a run that
applied every update: `requirements_updated: true, warnings: [], has_warnings:
false`, file byte-for-byte unchanged.
Four defects on that path, all long-standing (traced to 2bc295b32d, v1.2.0):
1. The only cross-check compared REQUIREMENTS.md's own body against its own
Traceability table, so an ID cited by ROADMAP but defined in neither was
invisible to it. `citedReqIds` is now hoisted out of the `if (reqMatch)` block
and cross-checked; ghost IDs raise a warning through the existing
warnings/has_warnings surface that execute-phase.md already prints.
2. `if (reqUpdate.ok)` had no `else`, so a Traceability-row write matching
nothing was discarded silently. Misses are now recorded.
3. `requirements_updated` was set unconditionally inside `if (existsSync)`,
reporting "the file was in the transaction" rather than "a write landed". It
now reflects whether the content actually changed.
4. DEFERRED_HEADING_RE's bare `v\d+` alternative treated an active `## v1
Requirements` heading as deferred, zeroing the body scan for that section.
Dropped; `deferred|backlog|future` still match, so #1159's suite still holds.
Also fixes the same class in gap-checker: the `items.length === 0` early return
fired before ghost rows were folded in, so an all-unregistered phase reported
LESS than a partially-unregistered one. It is now gated on ghostReqIds too.
Follows the #2140 precedent in milestone.cts, which hardened this exact class for
`requirements mark-complete`; adapted to this function's single `warnings` array
rather than adding fields execute-phase.md never reads.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SLufH5sDuqA1AiEGu45cuA
* fix(#2316): address review — milestone-aware deferred headings, probe-based ghost check
Independent review found the first pass green but wrong — the tests were built to
miss the shape that breaks.
BLOCKER: dropping the `v\d+` alternative from DEFERRED_HEADING_RE regressed #1159
against GSD's OWN shipped template. gsd-core/templates/requirements.md:35 ships
`## v2 Requirements` with the deferred-ness in the BODY PROSE ("Deferred to future
release"), not the heading — so `v\d+` was the only alternative matching it.
Deleting it made every scaffolded project emit a false traceability warning on
every phase complete, forever. The real problem is that `v\d+` matched BOTH the
active v1 and the deferred v2: deleting it swings from suppress-everything to
suppress-nothing. The heading check is now milestone-aware — a `## v<N>` heading
is deferred only when <N> is not the current milestone's major, resolved through
the existing state.cjs seam. Unresolvable milestone fails safe to the old
always-deferred behavior, since suppressing a warning beats spamming every project.
HIGH: the ghost check re-derived membership from two lossy index sets that
disagreed with the case-insensitive write paths, so it warned "not registered
anywhere" about an ID whose checkbox it had just ticked, and about a
case-mismatched ID whose write landed. It now probes the same two write surfaces
the writes use, mirroring milestone.cts's #2140 precedent (probe the write, don't
re-derive it) — which is what the original brief asked for and the first pass
didn't do.
HIGH: the cited-ID tokenizer split the rest of the line on whitespace, so every
trailing word became a "cited REQ-ID". The shipped roadmap.md template puts an
inline HTML comment on that exact line, so a fully correct run told users to
register `<!--` and `-->` as requirements. Cited IDs are now filtered to the
REQ-ID shape, matching what bodyReqIds/tableReqIds already require.
HIGH: gap-checker had the identical early-return defect 34 lines above the one
fixed — the could-not-parse branch gated ghost rows behind the same unguarded
items.length. One malformed <decisions> line made all ghost rows vanish.
Tests: the #2316-5 guard only covered `## Deferred`/`## Backlog`/`## Future` —
headings that trivially still match — omitting `## v2 Requirements`, the only
shape the change altered. Added the shipped-template fixture plus guards for each
finding above.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SLufH5sDuqA1AiEGu45cuA
* docs(#2316): backfill PR number 2339 into changeset
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SLufH5sDuqA1AiEGu45cuA
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GSD Core
Git. Ship. Done.
English · Português · 简体中文 · 日本語 · 한국어
A light-weight meta-prompting, context engineering, and spec-driven development system for Claude Code, OpenCode, Antigravity CLI, Kimi CLI, Kilo, Codex, Copilot, Cursor, Windsurf, and more.
What is GSD Core
GSD Core is a context-engineering and spec-driven development framework that drives AI coding agents (Claude Code, Codex, Antigravity CLI, Kimi CLI, Copilot, Cursor, and more) through a disciplined phase loop. It solves context rot — the quality degradation that accumulates as an AI fills its context window — by running all heavy research, planning, and execution work in fresh-context subagents while keeping your main session lean.
How it works
Each milestone repeats the same five-step loop, one phase at a time:
- Discuss — capture implementation decisions before anything is planned
- Plan — research, decompose, and verify the plan fits a fresh context window
- Execute — run plans in parallel waves; each executor starts with a clean 200k-token context
- Verify — walk through what was built; diagnose and fix before declaring done
- Ship — create the PR, archive the phase, repeat for the next one
Quickstart
npx @opengsd/gsd-core@latest
The installer prompts for your runtime (Claude Code, OpenCode, Antigravity CLI, Kimi CLI, Kilo, Codex, Copilot, Cursor, Windsurf, and more) and whether to install globally or locally. The installer is required for cross-runtime compatibility — do not copy files from agents/ or commands/ directly.
On another runtime or without Node.js? See Install on your runtime.
Once installed, start a new project or onboard an existing repo:
/gsd-new-project # greenfield project
/gsd-onboard # existing codebase
New here? Follow Your first project for a guided walkthrough from install to first shipped phase, or Onboarding an existing codebase for brownfield setup.
Documentation
What's new in 1.7.0 → docs/whats-new-1.7.0.md
Tutorials — learning by doing:
How-to guides — task-focused recipes:
Reference — authoritative facts:
Explanation — concepts and design decisions:
Full index: docs/README.md. Other languages: 日本語 · 한국어 · Português · 简体中文.
Why it works
Most AI-coding setups fail at scale because context bloat silently degrades output quality, there is no shared memory between sessions, and nothing verifies that code actually works. GSD Core solves all three: heavy work runs in fresh subagents, structured artifacts like STATE.md and CONTEXT.md survive session boundaries, and the verify step walks through what was built and generates fix plans before a phase is declared done. See docs/explanation/context-engineering.md for the full reasoning.
Troubleshooting? See docs/how-to/recover-and-troubleshoot.md.
Community
| Project | Platform |
|---|---|
| gsd-opencode | Original OpenCode port |
| Discord | Community support |
Star History
License
MIT License. See LICENSE for details.
Claude Code is powerful. GSD Core makes it reliable.