feat: Agent Skills subsystem (SKILL.md, autonomous invocation, commands via ACP/MCP/CLI) - #191

Merged

EvilFreelancer merged 14 commits into

mainfrom

Jul 15, 2026

Commits on Jul 14, 2026

docs(skills): design spec for skills subsystem

EvilFreelancer
and
claude
committed
feat(skills): Layer 1 — Skill model and SKILL.md loader

EvilFreelancer
and
claude
committed
feat(skills): Layer 2 — SkillRegistry and SkillsConfig

EvilFreelancer
and
claude
committed
feat(skills): Layer 3a — auto-register skills into the system prompt

EvilFreelancer
and
claude
committed
feat(skills): Layer 3b — use_skill tool + AgentContext.available_skills

EvilFreelancer
and
claude
committed
feat(skills): Layer 4 — wire skills through config, BaseAgent, AgentFactory

EvilFreelancer
and
claude
committed
feat(skills): Layer 5 (ACP) — skills as available_commands + slash expansion

EvilFreelancer
and
claude
committed
feat(skills): Layer 5 (MCP) — expose skills as MCP prompts (slash commands)

EvilFreelancer
and
claude
committed
feat(skills): Layer 6 — CLI --list-skills and GET /v1/skills

EvilFreelancer
and
claude
committed
test(skills): e2e autonomous invocation + example skills

EvilFreelancer
and
claude
committed
docs(skills): framework docs (en+ru), nav, config example + linter pass

EvilFreelancer
and
claude
committed
fix(skills): address subagent review (codex + claude)

EvilFreelancer
and
claude
committed
refactor(skills): address review — BaseSkill, drop MCP prompts, default .agent/skills

EvilFreelancer
and
claude
committed
fix(skills): /skill-name injection + use_skill gate; fix sed-corrupted strings

EvilFreelancer
and
claude
committed