feat(skills): 接入技能使用埋点,修复活跃技能被 curator 误归档

技能正文不进 system prompt,Agent 按 Progressive Loading 用 read_file
加载 SKILL.md 才算真正用到该技能。此前 read_file 与 skill_manage 的
patch/edit 都不更新 .usage.json,curator 的年龄流转据 latest_activity_at
判断闲置,导致每天都在用的 agent 创建技能在 30/90 天后被误标陈旧并归档。

- sandbox/tools.py: 新增 _custom_skill_name_from_path / _record_skill_read,
  read_file 读取 custom 技能目录下文件时 bump_use(best-effort,失败不影响主流程)
- skill_manage_tool.py: patch/edit 经 _bump_patch_best_effort 记 patch_count
- tests/test_skill_usage_signals.py: 12 项,覆盖路径解析、埋点与异常兜底

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
xuhdandClaude Opus 4.7 committed 2026-05-18 10:22:15 +08:00
1 parent 39def1c20b
commit 0386f54449
2 files changed
+57

No files matched your search

@@ -66,6 +66,21 @@ async def _to_thread(func, /, *args, **kwargs):
return await asyncio.to_thread(func, *args, **kwargs)
async def _bump_patch_best_effort(name: str) -> None:
"""记一次技能维护活动(patch / edit)。
刷新 ``.usage.json`` 的 ``patch_count`` 与 ``last_patched_at``,让 curator 的
年龄流转把"Agent 正在主动维护的技能"算作活跃,而不是误判为闲置后归档。
best-effort:埋点失败只记 DEBUG 日志,绝不影响技能写入主流程。
"""
try:
from deerflow.skills.usage import bump_patch
await _to_thread(bump_patch, name)
except Exception:
logger.debug("Failed to bump patch count for %r", name, exc_info=True)
async def _ensure_skill_db_registered(store: SkillStore, name: str, user_id: str) -> None:
"""Idempotently record an agent-created skill's ownership in the DB.
@@ -229,6 +244,7 @@ async def _skill_manage_impl(
name,
_history_record(action="edit", file_path=SKILL_MD_FILE, prev_content=prev_content, new_content=content, thread_id=thread_id, scanner=scan),
)
await _bump_patch_best_effort(name)
await refresh_skills_system_prompt_cache_async()
return f"Updated custom skill '{name}'."
@@ -253,6 +269,7 @@ async def _skill_manage_impl(
name,
_history_record(action="patch", file_path=SKILL_MD_FILE, prev_content=prev_content, new_content=new_content, thread_id=thread_id, scanner=scan),
)
await _bump_patch_best_effort(name)
await refresh_skills_system_prompt_cache_async()
return f"Patched custom skill '{name}' ({replacement_count} replacement(s) applied, {occurrences} match(es) found)."