mnemos-brieflisted
Install: claude install-skill klarlabs-studio/mnemos
# Mnemos Brief
Manual counterpart to the `SessionStart` brief hook. The hook fires once, at
session start, and emits two lines. This skill is the deep version: run it any
time the user wants to know what the brain actually holds — mid-session, after
a `/clear`, or when the automatic brief never fired.
## Choosing a path
Prefer the **MCP tools** (`mcp__mnemos__*`). They work identically against a
local brain and a hosted one, and they return structured results.
Fall back to the **CLI** only when the MCP server is not connected — i.e. no
`mcp__mnemos__*` tool is available, or every call errors. Do not shell out to
`mnemos` when the MCP tools are working; the CLI opens its own connection to
the brain and may resolve a different DSN than the MCP server is serving.
## Procedure — MCP path
Run these concurrently where possible; each is independent.
1. `knowledge_metrics` — total claims, runs, open contradictions. This is the
headline. If it fails, the brain is unreachable: say so and stop.
2. `query_knowledge` with a question drawn from the user's current work (the
repo name, the branch, the feature under discussion). If the user gave no
topic, use the repository name from the working directory.
3. `list_dissonances` — open contradictions the agent is expected to
reconcile. Cap what you report at the 5 most relevant; say how many were
omitted rather than silently truncating.
4. `list_decisions` — the most recent recorded decisions, for continuity with
the l