| name | rtime-context |
| description | Use when planning dynamic context unlocking, building ContextUnlockPlan or Context Pack skeletons, explaining why context should be loaded, routing requests across brain/rtime-hub/runtime evidence, or designing why-context behavior without reading sensitive data. |
Rtime Context
Use this skill for read-only context planning before retrieval or action. The
stable command surface is the rtime-context CLI package in
packages/rtime-context.
First Reads
From <repo-root> on Mac or
<repo-root> on orangepi, read:
AGENTS.md
docs/context-unlocking.md
docs/context-orchestrator.md
docs/prompt-layering.md
docs/tooling-packaging.md
Rules
- Keep context planning read-only.
- Treat the output as a plan, not proof that context has already been loaded.
- Do not read secret file bodies, identity data, addresses, session stores, or
credentials from this tool.
- Use
rtime-context plan before broad cross-project retrieval when the task
could involve brain, rtime-hub, runtime logs, or sensitive context.
- Use
rtime-context explain when the user asks why a source or memory should
be unlocked.
- Context Pack skeletons may list sources and recommended tools, but actual
source reads still need task relevance and normal validation.
Commands
Install locally when needed:
scripts/install-rtime-tooling.sh --tool rtime-context --profile mac
python -m pip install -e packages/rtime-context
rtime-context doctor
Read-only checks:
rtime-context plan "检查 rtime-hub 项目状态和 brain 文献引用"
rtime-context pack "重构 Feishu bridge 并跑 pytest" --workspace <repo-root>
rtime-context explain "为什么要读取 runtime logs?"
printf '{"jsonrpc":"2.0","id":1,"method":"tools/list"}\n' | rtime-context-mcp
Validation
For package, MCP, skill, or plugin changes, run:
python -m py_compile packages/rtime-context/src/rtime_context/*.py
PYTHONPATH=packages/rtime-context/src python -m pytest tests/test_rtime_context_cli.py tests/test_rtime_context_mcp.py -q
scripts/validate-codex-plugin.py plugins/rtime-context
git diff --check