[shepherding] Pi five-role skill capture pipeline + `/tdai-memory-sync-skills`

Is this feature related to a problem? | 该功能需求是否与某个问题相关？请描述

Shepherding issue split out of #1070. Follow-up PR author: @kbob3687-hub. Targets feat/server_team.

Background

Current state: SkillClient.exportSkill and skill-core can round-trip skills, but there's no closed loop that pulls mined SKILL.md back into Pi's native skills directory without overwriting user-authored ones. #1070's five-role capture pipeline is the missing half of that loop and is the shape we'll eventually want on other clients too.

In scope

• Five-role capture pipeline as landed in #1070 (roles + priorities documented in the PR description)
• /tdai-memory-sync-skills command that pulls mined SKILL.md and merges into Pi native skills
• Non-overwrite guarantee for user-authored skills — a mined skill MUST NOT clobber a hand-written one silently
• Wire-in point: Pi adapter command surface + existing skill export APIs

Out of scope

• New skill mining algorithms (reuse what MemoryCore already has)
• Exposing skills through MemoryKnowledge MCP — that's #833 Part D and is being split into its own issue
• Non-Pi clients

Acceptance criteria

• /tdai-memory-sync-skills pulls skills scoped to the current agent/team only (no cross-tenant leakage)
• User-authored SKILL.md files are detected and preserved (test: place a hand-written skill, run sync, assert file unchanged)
• Conflict resolution documented: what happens when a mined skill and a user-authored skill share a name
• Sync is idempotent (running twice ≠ duplicating skills)
• No modifications to handler.ts or session/codebuddy/init.ts
• Docs describe the five roles and their priorities

Related

• Closes down the skill capture portion of #1070
• Downstream of SkillClient.exportSkill — no new export primitives
• Adjacent to #833 Part D (MemoryKnowledge MCP skill tools) but does not depend on it

Suggested branching

Branch name: pi-adapter/skill-capture-pipeline against feat/server_team.

Describe the solution you'd like | 描述你期望的解决方案

/

Describe alternatives you've considered | 描述你考虑过的其他方案

No response

Additional context | 补充说明

No response