Harness-boundary enforcement for open-gsd/gsd-core contributions: Claude Code PreToolUse hooks that DENY a broken issue, PR, or push before it happens — plus the contribution and maintainer-review skills behind them. Reuses gsd-core's LIVE gate scripts instead of reimplementing its policy.

agent-toolingai-agentsclaude-codecode-reviewcontribution-workflow

gsd-contribution-toolkit

Repo · 1 star · JavaScript

▲ 0

A self-contained, toggleable GSD capability that enforces clean open-gsd/gsd-core contributions at the Claude Code harness boundary — 12 fail-closed PreToolUse gates + 2 skills + 5 commands, installed via gsd-core's capability system. Advisory-only on non-Claude runtimes.

capabilityclaude-codecode-qualitydeveloper-toolsgit-hooks

gsd-handover

Repo · 0 stars · JavaScript

▲ 0

Pattern B session handover for GSD on Claude Code: pass a baton to a fresh session before a context or usage limit kills a phase mid-run (BETA, default-off).

gsd-messaging

Repo · 0 stars · JavaScript

▲ 0

Thin, default-off, Claude-Code-only GSD capability: live inter-agent escalation, agent interview, and a learning doorbell over Claude Code's shipped messaging. Sibling of gsd-handover.

verifier-reach-is-spec-reach

Repo · 1 star · JavaScript

▲ 0

Verifier Reach Is Spec Reach — paper + reproducible experiments: an LLM code-verifier can only catch defects the spec named.