跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
效率与协作

commitment-tracker

Track commitments in Littlebird meeting records. Use to see what you owe, what others owe, and what has gone stale. Read README.md for the guide map.

Mario Aldayuz / Littlebird · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add legioncodeinc/vibe-coding-tools --skill commitment-tracker

技能说明


name: commitment-tracker description: "Track commitments in Littlebird meeting records. Use to see what you owe, what others owe, and what has gone stale. Read README.md for the guide map." license: AGPL-3.0-or-later compatibility: Claude Cowork, Claude Code 2.1 or newer, Cursor 2.4 or newer, Codex metadata: hive-tier: standalone version: "1.0.0" author: "Mario Aldayuz / Littlebird" requires: "Littlebird MCP (Power or Pro plan)"

Commitment tracker

Purpose

Every Littlebird meeting summary already ships a structured ## Action Items block with an owner tag on each line, a ## Decisions block with a decider tag, a ## Risks / Open Questions block, and a ## For You section. That structure sits there unharvested. In the account this skill was validated against, one 30 minute partnership call produced four action items, two of them tagged Unassigned, plus two open questions, and none of it went anywhere.

This skill does three things with that structure:

  1. Harvests it across every recorded meeting in a window.
  2. Verifies it, which is the part nobody does: searches for downstream evidence that the thing actually happened.
  3. Escalates by age, changing channel and framing rather than repeating itself.

It does not re-extract commitments from raw transcript. Raw transcript chunks are weakly diarized and frequently tagged [Others], and state of the art diarization runs at 11 to 13 percent error (references/research/distilled-commitment-tracking.md, section 4). Attribution comes from the summary blocks. Always.

Capability gate

This skill requires the Littlebird MCP on a Power or Pro plan.

Before anything else:

  1. List the tools actually available in this session and use the real tool names. Do not assume a tool exists because it is named in references/littlebird-mcp-reference.md.
  2. If no Littlebird MCP tools are present, stop and tell the user the skill needs the Littlebird MCP connected. Do not attempt a partial run from memory or from other sources.
  3. If routine creation is part of the request, call LB_INTERNAL_GET_SUBSCRIPTION_STATUS first to confirm the plan supports another routine.

Read references/evidence-standards.md before writing any output. Every line in the ledger is observed, inferred, external, or unknown, and the kind is visible to the reader.

Littlebird MCP calls used

ToolUsed for
LB_INTERNAL_LIST_MEETINGSEnumerating the window with start_date, end_date and a limit high enough to cover it. Also prior instances of a recurring meeting by name. Entries carrying an id are recorded; entries without one are unrecorded calendar events and are not searchable. Name lookup uses this tool.
LB_INTERNAL_GET_MEETINGThe harvest, on every recorded id: ## Action Items with the owner tag on each line, ## Decisions with the decider tag, ## Risks / Open Questions, and ## For You.
LB_INTERNAL_SEARCH_MEETINGSWhether a later meeting confirmed delivery. query is the deliverable named as a noun, start_date the day after the origin meeting. Topic lookup uses this tool. Do not add attendees; that filter is OR and best-effort and can miss a matching meeting entirely.
search_user_contextThe downstream evidence sweep and the deliberate absence proof. Narrow parallel search_queries and search_queries_messages, a standalone_query, a date_range starting the day after the commitment, and filters on data_source of messages, snapshots or summaries, or on app when proving absence.
LB_INTERNAL_GET_MEETING_TRANSCRIPTAlmost never. The single narrow case is a decision's supporting quote, taken for wording only (references/decisions-log.md). Never for attribution, never in the harvest.
LB_INTERNAL_GET_ROUTINE_REPORTSRead first on every run, limit 5. Supplies the carry-forward list, the restatement counts that drive escalation, and everything the user already closed or held.
LB_INTERNAL_GET_SUBSCRIPTION_STATUSConfirming the plan supports another routine, before CREATE_ROUTINE.
LB_INTERNAL_CREATE_ROUTINEStanding up the weekly ledger routine, on user approval. Works from an interactive session. Blocked only from inside a running routine.
LB_INTERNAL_GET_ROUTINE_CONFIG and LB_INTERNAL_UPDATE_ROUTINEEditing that routine later. UPDATE replaces the whole prompt and the whole schedule, so always GET first.

There is no Littlebird tool that searches past Littlebird chat conversations. Where a message history is needed, search_user_context with data_source: messages is the tool.

Exact call shapes, with the parameters to substitute: the retrieval brief below.

Trigger

Ask for it: "what did I promise", "what do people owe me", "who owes me what", "did I follow up", "open action items", "what fell through the cracks", "weekly commitment review", "chase my follow-ups".

Scheduled: the weekly routine below, whose report ends by naming this skill for the deep run.

Do not trigger for: the follow-up message after one call (that is meeting-scribe), or an unanswered thread where nothing was promised (that is who-am-i-ghosting).

Routine cadence

Weekly. Default Monday 08:00 local, the schedule in the routine wiring section below.

Offer to create it from this session with LB_INTERNAL_CREATE_ROUTINE: show the user the exact prompt text and the schedule, get approval through AskUserQuestion, then call it. Do not tell the user to go set it up by hand.

ModeTriggerWindowOutput
Weekly routineScheduled, unattendedLast 7 days plus open carry-forwardA routine report, no files, no approvals
Deep runUser asks in Cowork or Claude CodeLast 30 days, or 90 on a first backfillThe full ledger file plus drafted nudges held for approval

The routine observes. The deep run acts. A routine cannot create or update routines, and cannot hold an app

阅读完整来源 ↗