Turn a finished research session into a file knowledge-intake can read
without reformatting.
When To Use
After a research session has findings and you want them stored rather than merely reported.
When NOT To Use
- The session is still gathering findings: finish
Skill(tome:research)first, since an export of an empty session emits only frontmatter. - You want a human-readable report: use
Skill(tome:synthesize), which produces the narrative form with coverage and gaps.
Workflow
Load the session and render it:
from pathlib import Path
from tome.session import SessionManager
from tome.output.export import export_for_memory_palace
session = SessionManager(Path.cwd()).load_latest()
markdown = export_for_memory_palace(session)
Path(f"docs/research/export-{session.id}.md").write_text(markdown)
SessionManager resolves its store relative to the path it is given, so
run this from the repository root or the session will not be found.
To export a specific session, use load(session_id) instead of
load_latest().
What It Emits
YAML frontmatter carrying topic, domain, session_id, date,
finding_count, channels, and type: research-export, followed by
findings grouped by channel. Each finding renders its source, URL,
relevance, summary, and channel-specific metadata.
An empty session emits the frontmatter and a line stating that no findings were recorded, which is deliberate. A silent empty file would read as a successful export of nothing.
Handoff
The output is an external resource, not a corpus entry. Feed it to
Skill(memory-palace:knowledge-intake), which owns evaluation, routing,
and storage. This skill does no scoring.
Exit Criteria
-
export_for_memory_palacewas called on a loaded session - The output file exists and its frontmatter includes
type: research-export -
finding_countin the frontmatter matches the session's findings - The file was handed to
Skill(memory-palace:knowledge-intake)or its path was reported for later intake