brief

name	brief
description	Session-start briefing from Wenlan for Codex. Reads the project status file, loads identity/preferences/topic memories, and surfaces pending revisions. Invoked as /brief [topic].
argument-hint	[topic]
allowed-tools	["Bash","mcp__wenlan__context","mcp__wenlan__recall","mcp__wenlan__list_pending_revisions","mcp__wenlan__accept_revision","mcp__wenlan__dismiss_revision"]
user-invocable	true

/brief

Pull a curated session brief from Wenlan. Use it at session start or after a major topic shift.

Sources, in order:

1. Project status file: what /handoff last wrote.
2. context MCP: identity, preferences, and topic-relevant memories.
3. list_pending_revisions: daemon-flagged memories awaiting human review.

The status file wins for "what is next" because it is the live ledger.

1. Read project status first

Detect project root:

cd_repo="$(git -C "$PWD" rev-parse --show-toplevel 2>/dev/null || true)"
if [ -n "$cd_repo" ]; then
  project="$(basename "$cd_repo")"
else
  project="$(basename "$PWD")"
fi
printf '%s\n' "$project"

Read ~/.wenlan/sessions/_status/<project>.md:

cat "$HOME/.wenlan/sessions/_status/<project>.md"

If the file exists, render its ## Last session, ## Active, and ## Backlog sections verbatim at the top under Status. If the file is missing, say nothing about it.

2. Resolve topic and space

Accept an optional topic argument and an optional inline space:<name> token. Extract the space token before using the rest as topic text:

raw_args="<the full argument string passed to /brief>"
space_arg="$(printf '%s\n' "$raw_args" | grep -oE 'space:[A-Za-z0-9_-]+' | head -1 | cut -d: -f2)"
topic_arg="$(printf '%s\n' "$raw_args" | sed -E 's/[[:space:]]*space:[A-Za-z0-9_-]+[[:space:]]*/ /g' | sed -E 's/^[[:space:]]+|[[:space:]]+$//g')"

Call the Codex resolver:

resolved="$(plugin-codex/bin/resolve-space.sh --cwd "$PWD" ${space_arg:+--arg "$space_arg"} ${topic_arg:+--topic "$topic_arg"} 2>/dev/null)"
space="$(printf '%s\n' "$resolved" | cut -f1)"
source_layer="$(printf '%s\n' "$resolved" | cut -f2)"

If space is non-empty, print Resolved space: <space> (from <source-layer>) and pass it to the context MCP call. If it is empty, print Resolved space: none (unscoped) and omit the space parameter.

3. Call context

Call the Wenlan MCP context tool:

context(topic="<topic_arg or inferred topic>", space="<resolved if non-empty>")

If the user omitted a topic, infer it from the working directory and recent conversation. Do not ask unless inference would be misleading.

Use the result to model how the user thinks, not just to retrieve facts. Their preferences, corrections, and past decisions shape how you should work.

4. Pending revisions check

After loading context, call:

list_pending_revisions(limit=10)

If the result is empty, say nothing. If the call errors, print one warning and continue.

If non-empty, show the top three:

Pending revisions (<N> total, top 3 shown):

1. target: <target id>  (proposed by <source_agent or "daemon">)
   revision: "<revision text>"
   Action: accept (replace original) | dismiss (drop revision) | skip

Inline verbs map to:

• accept: accept_revision(target_source_id="<id>")
• dismiss: dismiss_revision(target_source_id="<id>")
• skip: no call

If there are more than three, end with:

Run /curate revisions to walk the full queue.

Do not auto-action revisions.