fix: MCP工具缺乏对禁用CLI任务的替代指引与能力映射

[binggg]

Attribution issue

• issueId: issue_mo8xicmr_y3lqj3
• category: tool
• canonicalTitle: MCP工具缺乏对禁用CLI任务的替代指引与能力映射
• representativeRun: atomic-js-cloudbase-cli-log-query-by-request-id/2026-04-21T17-58-43-zmic3p

Automation summary

---

SUMMARY:

• root_cause: When the runtime has CloudBase CLI disabled but MCP tools enabled, and the task explicitly asks to "use CloudBase CLI", the agent has no guidance on how to map CLI commands to equivalent MCP tools. The runtime capability notice clearly states CLI is disabled, but without explicit capability mapping guidance, agents don't know which MCP tools to use as alternatives.
• changes: Added CLI-to-MCP capability mapping guidance to two skill files:

1. config/source/skills/cloud-functions/SKILL.md (lines 239-273): Added "When CLI is disabled but MCP is available" section with a detailed mapping table translating common CLI commands (tcb fn list, tcb fn deploy, tcb fn invoke, tcb fn log, etc.) to their equivalent MCP tool actions. Also added invokeFunction and queryLogs references to the relevant tool sections.
2. config/source/skills/cloudbase-cli/SKILL.md (lines 60-70): Added "When CLI is disabled" section that explicitly tells agents NOT to attempt CLI commands when disabled, and points them to the cloud-functions skill for the capability mapping.

• validation: All 3 regression tests pass (10 tests total):
• `build-skills-repo.te

Changed files

• config/source/skills/cloud-functions/SKILL.md
• config/source/skills/cloudbase-cli/SKILL.md

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[binggg]

Attribution post-PR evaluation

• visibility: internal identifiers, run ids, and private links are intentionally omitted
• attempt: 1
• eval_scope: primary_only
• overall: FAILED
• summary: at least one planned evaluation case failed
• updated_at: 2026-04-28T21:47:41.716Z

Cases

• [FAILED] — primary — evaluation failed