harness SDD suite PR brings spec-driven dev, dispatch discipline, adversarial review to anthropics/skills
jordimarsal · ghdev · 2026-10-16
Developer jordimarsal opened PR #2039 on anthropics/skills adding three process-only skills distilled from harness (MIT), a multi-agent spec-driven development harness battle-tested on real projects (an Astro portfolio site, a Kafka telemetry pipeline, a Java/Spring Boot status API):
- spec-driven-development: EARS-formatted requirements, 3-file specs (requirements/design/tasks), human approval gates, plus a mechanical requirement-to-test traceability checker (checktraceability.py) enforcing that every R<n> maps to a real test, with exit 0/1/2 and a JSON protocol.
- agent-dispatch-discipline: subagent orchestration protocol — one dispatch = one batch of 2-4 tasks (never 5+), on-disk evidence checks after every batch, and fresh-session restarts at 70% context.
- adversarial-review-gauntlet: sequential review G1 correctness → G2 adversarial → G3 maintainability → G4 evidence, with gauntlet.sh as a mechanical G1 gate; a REJECT stops everything.
The skills are self-contained (no API keys or binaries) and validated: each SKILL.md under 500 lines, agentskills-spec compliant, and gate scripts pass frozen PASS/FAIL fixtures.
More from coding & agent
- Index Network open-sources Social Discovery Protocol: agents negotiate fit on your behalf — carsonfarmer · 2026-10-23
- Installed a formal-verification skill: it silently installed Java, MacBook runs 10 degrees hotter — zealcaiden · 2026-10-23
- OpenAI's Decisions API hits GA: near real-time routing up to 10x faster than GPT-6 Luna — OpenAIDevs · 2026-10-23
- Ruff author warns: upgrade Node/Bun now — --use-system-ca still accepts the eMudhra cert — charliermarsh · 2026-10-23
- Proof of Agent: paste one prompt and your AI mines, grades, and self-governs — adamamcbride · 2026-10-23
- DeepLearning.AI weekly: GPT-6.1 hits 97.9% byte parity on COBOL migration — DeepLearningAI · 2026-10-23