feat: semport skill -- semantically-aware cross-tool and cross-language code porting

[AlexMikhalev]

Summary

Add a semport skill to terraphim-skills that teaches agents how to perform semantically-aware code ports -- preserving intent and architecture rather than mechanical line-by-line translation.

Based on StrongDM's Factory technique: https://factory.strongdm.ai/techniques/semport

Three Port Models

Model	Use Case
Cross-tool	Port SKILL.md + agent prompts from Claude Code to OpenCode, Codex, Gemini CLI, Copilot, Kiro
Cross-language	Evaluate upstream commits and translate with semantic adaptation (Python -> Rust)
Adaptive	Reshape external API surface into internal conventions while maintaining semantic equivalence

Skill Structure

skills/semport/
  SKILL.md                        -- technique, workflows, constraint rules
  references/
    tool-format-map.md            -- Claude Code format -> each target tool format
    validation.md                 -- port correctness validation strategies
    examples.md                   -- concrete before/after per port type

Integration Points

• terraphim-hooks KG: source->target term mappings for format conversion tokens
• terraphim-agent replace: mechanical token replacement layer beneath semantic translation
• Judge system: validate port quality via behavioral test comparison
• Compound engineering P5: this skill IS the implementation of cross-tool conversion

Implementation Plan

See: plans/semport-terraphim-skill.md in cto-executive-system

Steps:

1. Create SKILL.md with three-model workflow and constraint hierarchy (~1h)
2. Create tool-format-map.md reference mapping Claude Code -> 6 targets (~1h)
3. Create validation.md with per-port-type validation strategies (~30m)
4. Create examples.md with concrete before/after ports (~30m)
5. Wire into compound-engineering adoption plan P5 (~15m)

Estimated effort: ~3.5h

Acceptance Criteria

• SKILL.md covers all three port models with decision tree
• tool-format-map.md maps Claude Code skill format to all 6 target tools
• At least one concrete example per port model in examples.md
• Validation strategy defined for each port type
• Cross-referenced with compound-engineering adoption plan P5

Related

• Compound engineering adoption plan Priority 5 (cross-tool conversion)
• feat: implement .well-known/skills/ discovery endpoint (Cloudflare RFC) #57 (.well-known/skills/ discovery -- semport skills would be discoverable)
• Every's compound-engineering-plugin converter CLI (one-time semport; this skill adds ongoing/adaptive)