Rehydrate a fresh session from on-disk state.
Install via CLI
openskills install gsaranti/pantheon---
name: metis-session-start
description: Rehydrate a fresh session from on-disk state.
disable-model-invocation: true
---
# /metis-session-start
Read the minimum set of files that orients the agent to the project's current state.
## Preconditions
`.metis/` must exist. If it does not, stop and point at `/metis-init`.
Partial state inside `.metis/` (missing or incomplete files) surfaces as an Anomaly in the Return rather than a hard error.
## Load
In this order:
1. **`.metis/CURRENT.md`** — the previous session's handoff. The primary rehydration source.
2. **`.metis/BUILD.md`** — the architecture brief, when it exists. Loaded once at session start so the rest of the session's work can lean on it without re-reading.
3. **`.metis/INDEX.md`** — the concept → source-doc map, when it exists.
## Do not load
- `.metis/SYNTHESIS.md`, `.metis/CONTRADICTIONS.md`, `.metis/QUESTIONS.md`, `.metis/RESOLVED.md` — reconcile artifacts. Loaded on demand by the skills that need them, not at session start.
- `docs/` — source corpus. Loaded on demand by skills that turn on specific passages.
## Write scope
**None.**
## Return
- **What happened last session** — one paragraph from `CURRENT.md` *What happened*. If `CURRENT.md` is missing or empty, say so directly.
- **In flight** — what's being worked on right now, from `CURRENT.md` *Current state*. If nothing is in flight, say so.
- **Open questions** — the list from `CURRENT.md` *Open questions*. Surfaced as items the user may want to address.
- **Where to start** — directly from `CURRENT.md` *Where to start*. Do not rewrite — pass it through.
- **Anomalies** — anything unexpected: missing `CURRENT.md` (suggest `/metis-init`), `CURRENT.md` referencing files that don't exist on disk, `.metis/BUILD.md` mentioned by the handoff but absent. Surface rather than absorb.
## Invocation prompt
Silently accept and ignore any trailing free-text prompt.
No comments yet. Be the first to comment!
Implement persistent memory patterns for AI agents using AgentDB. Includes session memory, long-term storage, pattern learning, and context management. Use when building stateful agents, chat systems, or intelligent assistants.