mirror of
https://github.com/akitaonrails/ai-memory.git
synced 2026-10-02 03:24:46 +08:00
Lucas Oliveira
6f9c894020
Mirror the native gate in hooks/kimi-code/user-prompt-submit.sh/.ps1: the first user prompt of a session fetches /handoff with the [briefing] params, later prompts fetch it without (handoff fetch itself stays — cheap and self-limiting). _lib.sh gains ai_memory_parse_toml_flag (bare or quoted values, parity with hook_capture.rs), ai_memory_briefing_qs and ai_memory_briefed_file without touching ai_memory_marker_qs (other agents unchanged); the PowerShell lib gains the equivalents behind a -BriefingOncePerSession switch. State: <data dir>/briefed/<safe-key> marked only after the GET completes, success or error (fail-open). docs/marker-file.md and CHANGELOG note that for Kimi Code the brief is delivered on the first prompt of the session (kimi discards SessionStart hook stdout) and that re-briefing after /clear is not supported in v1.
10 lines
558 B
PowerShell
10 lines
558 B
PowerShell
# Kimi Code injects UserPromptSubmit hook stdout as a user message
|
|
# (origin hook_result) before the turn, so the pending handoff is
|
|
# fetched here. Empty stdout injects nothing. The compiled project
|
|
# brief ([briefing] inject_on_session_start) rides the FIRST prompt of
|
|
# the session only — kimi discards SessionStart hook stdout, so this
|
|
# is the once-per-session parity with Claude's SessionStart brief.
|
|
. "$PSScriptRoot\..\lib\ai-memory-hook.ps1"
|
|
Invoke-AiMemoryHook -Event "user-prompt" -Agent "kimi-code" -FetchHandoff -BriefingOncePerSession
|
|
exit 0
|