Terminator v0.2.0

Orchestration protocol refinement — this release upgrades the skill instructions themselves, based on a design review of how agents behave across small, large, and very different repositories.

Added

• Proportionality and scope rules: the protocol scales with the objective; small goals compress scanning, discovery, and planning into one pass, while verification and the final report are never skipped. Out-of-scope discoveries become recommendations, not actions.
• Plan-presentation rule: large, ambiguous, or high-impact objectives receive the orchestration plan before execution; small unambiguous objectives execute directly.
• Tool selection rules: confirm availability before planning around a tool, prefer dedicated over generic tools, keep discovery read-only, parallelize independent steps where supported, verify with deterministic checks.
• Report STATUS vocabulary: Complete, Partial, Blocked, or Ready — reports are now comparable across runs and environments.

Improved

• Verification is evidence-bound: success is never reported without at least one concrete check; unverifiable results are stated under LIMITS.
• Compact Tool Registry field names aligned with the tool discovery guide.
• Specification updated to match the refined protocol.

Install: copy skills/terminator/ into the skills folder used by your agent environment, then invoke with Use Terminator to <objective>.