Extract bounded, source-linked context from large files or many inputs. Use when context is too large, files are oversized, or a task needs concise excerpts.
中文简介正在准备中。当前显示原文说明。
Inspect SIPS Homebase status, manifest wiring, routes, host visibility, and MCP freshness. Use when asked for SIPS status, command center, host audit, plugin visibility, or whether Homebase is fresh.
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
Use the sips-homebase MCP first: call homebase_status for the status card, homebase_routes for route inventory, homebase_host_audit for Codex wiring, and homebase_mcp_freshness before claiming the live host is fresh.
Plugin MCP tools may be deferred. If Homebase is absent from the initial tool
list, use tool_search where the host exposes it and search for the exact
Homebase capability before falling back. Do not declare the MCP unavailable
until discovery returns no matching SIPS namespace or tool. Preserve the actual
native server/tool event when a discovered call succeeds.
Report the proof boundary explicitly: source/cache/config/child-process freshness is not the same as an already-open Codex session rediscovering the refreshed plugin.
If the MCP surface is unavailable, inspect .codex-plugin/plugin.json, .mcp.json, .agents/plugins/marketplace.json, state.yaml, and scripts/harness_homebase_mcp.py directly, then say the MCP path was unavailable.
If a diagnostic MCP JSON-RPC request is sent by piping initialize and
tools/call into python3 scripts/harness_homebase_mcp.py, label the result a
repo-local source subprocess. Preserve both layers: the inner source-subprocess
tools/call succeeded, while native task MCP callability remains unproven.
Report the observed outer host transport conditionally—for example, if the
pipeline was launched through exec, say the outer host call was exec, not a
native mcp__sips_homebase__* call. Do not say that the MCP tool was called
from the task, and do not include it in task_invoked_tools. That fallback does
not prove the installed cache, host configuration, task advertisement, native
callability, or persistent host attachment.
The read-only homebase_method exposes five advisory analyses. Persist a receipt through adaptation action analyze; transition authority remains in the controller. See method API and command protocol.
Use adaptation read views next and outcomes, and write actions advance and
link_outcome. Work packets are revision-bound. Review/stopped packets cannot
advance into activation. See ../../docs/sips-08.md for request contracts.
The adaptation dependencies read view explains current drift and its affected
receipts without modifying history. Registered sips.policy.v2 artifacts can be
tried with policy_trial and an explicit target_episode; the controller consumes
durable executed probe outcomes and pins their source revision. Missing evidence,
unavailable outcomes, and expired assumptions remain distinct. Freeze policy_cases
checks to evaluate behavior independently. Policy trials propose interventions;
normal lifecycle actions and exact-candidate activation remain authoritative.
Adaptation notice records evidence-linked skill/tool opportunities; opportunities
reads their source freshness and candidate handoff. next exposes the capture cue.
Use sips-in-run-improvement for deciding between creation, extension and repair.
See ../../docs/sips-in-run-improvement.md for fields and limits.
For episode progress, diagnosis, dependency drift or review explanations, read the
adaptation visual view and render its Mermaid diagram in the response. Use compact
charts for experiment comparisons, showing missing results as unknown and keeping
synthetic, foreground and production evidence distinct. Link to underlying receipts.
Do not substitute a chart for failures or limitations, or claim that returning HTML
or Mermaid source proves the host displayed it. The existing HTML widget renderer
uses a host-specific preview directive; use ordinary Mermaid or local image embeds
when that host integration is unavailable.
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Extract bounded, source-linked context from large files or many inputs. Use when context is too large, files are oversized, or a task needs concise excerpts.
中文简介正在准备中。当前显示原文说明。
Route a task through SIPS commands, agents, scripts, MCP tools, or bounded delegation. Use when asked to route, split, fan out, escalate, or choose the right SIPS path.
中文简介正在准备中。当前显示原文说明。
Turn failures, logs, symptoms, or flaky behavior into a compact repro and verification plan. Use when debugging needs a reproducible failure path.
中文简介正在准备中。当前显示原文说明。
Capture and develop a reusable skill, tool repair, or behavior improvement noticed while completing a task. Use when instructions are stale, a recurring workflow deserves a skill, or a tool contract fails in practice.
中文简介正在准备中。当前显示原文说明。
Search, inspect, and record SIPS-owned Memory Fabric lessons. Use when a task needs prior lessons, recurring-fix memory, recall health, scoped historical context, or when a just-fixed bump or error should be recorded.
中文简介正在准备中。当前显示原文说明。
Plan browser, app, screenshot, or UI checks before visual claims. Use when a task includes app shots, visual QA, generated assets, or runtime UI proof.
中文简介正在准备中。当前显示原文说明。