一键导入
follow-up-tracker
Log and surface pending commitments — things you owe others and things others owe you — from conversations, notes, and inbox snippets
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Log and surface pending commitments — things you owe others and things others owe you — from conversations, notes, and inbox snippets
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Host the memory-of-use ledger's schema/counting/status-vocabulary convention and the confirmed-proposal apply/verify/rollback/SECGATE machinery — the single, mandatory, deny-listed skill so this governing prose is reliably present from setup (Mode A + Mode B) and can never itself be an apply's target (ADR-061)
Host the skill-content pull flow — classify every installed curated skill via fresh-bytes-on-both-sides against the on-disk pool and the workspace's own install manifest, offer per-component updates with no silent overwrite, and stand as the mechanism that backfills the three mandatory safety skills (self-apply, self-archive, self-upgrade) into any workspace that runs this flow — Face 1 of the v2.19 Persistency Layer (KDQ-PULL), kept textually distinct from self-upgrade's engine-version Face 2 (ADR-072/073)
Host the auto-cleaning move-eligibility gate, destination gating, reversible-move-log rollback, and reference-integrity verification for proposing a stale/superseded file's relocation into the local archive convention — the sibling, mandatory, deny-listed skill to self-apply, holding the PATH-relocation operation type separately so neither module carries two verification concerns (ADR-066)
Host the kit-version walk-forward contract and the two-write-class self-integrity invariant (verify-then-swap on safety machinery) for replacing a workspace's own engine/framework machinery across kit versions — the third deny-listed sibling to self-apply/self-archive, dormant at v2.19 (zero real forward-walk targets) but installed and reachable now so no future rung has to backfill the gate (ADR-071)
[one sentence describing what this skill enables Cowork to do]
Run a periodic Collect, Process, Review, and Plan pass across workspace files to surface what's stalled, due, and worth prioritizing next week
| name | follow-up-tracker |
| description | Log and surface pending commitments — things you owe others and things others owe you — from conversations, notes, and inbox snippets |
| tools | ["claude-code"] |
| trigger_examples | ["Track follow-ups from this conversation","What do I owe people from this thread?","Pull pending commitments from these notes","What's outstanding from last week's meeting?"] |
Use follow-up-tracker when the user wants to extract and track two-directional commitments from a conversation, inbox thread, meeting notes, or any block of text containing next steps. The skill distinguishes between what the user owes others and what others owe the user, preserves ambiguity for uncertain items, and surfaces date hints. Use it after any interaction that generated commitments — before those commitments are forgotten or lost in context. This skill does NOT manage a task system; it extracts commitments from pasted content.
Plain markdown in the chat. Structure: ## What I owe (bullet list: commitment — to: [name] — due: [date or "no deadline set"]) then ## What others owe me (bullet list: commitment — from: [name] — due: [date or "no deadline set"]). Mark uncertain items with [uncertain]. No JSON, no YAML, no Obsidian wikilinks. Output is copy-paste-ready for a task app or notes folder.
[uncertain] with the original phrase.[uncertain] so the user can make the call.Input (meeting notes snippet):
April 30 — call with Sarah and Marcus.
- I said I'd send the revised contract draft by Thursday.
- Marcus mentioned he'd share the vendor comparison doc "sometime this week."
- Sarah asked if I could intro her to Jordan — I said I'd think about it.
- We should do a proper debrief on the Q1 numbers at some point.
Output:
Note: "Proper debrief on Q1 numbers" — informal suggestion only, no owner or date; excluded as not a commitment.
Follow-up-tracker is a structured extraction skill — most output is data-driven and profile-neutral (who, what, when). The section headers and inline notes are the only text where voice applies. Consult context/writing-profile.md only for the tone of inline notes (e.g., a formal profile produces "Item could not be confirmed as a commitment"; a direct profile produces "[uncertain]"). If no profile exists, default to terse and direct.