Action item audit

Most things that get dropped were never refused. They were agreed to in a meeting, written into notes nobody reopened, and never turned into tracked work. This skill finds them while they are still cheap to recover.

When to use this

• Before a retrospective — bring evidence instead of impressions
• Before a release — catch the commitment that was never scheduled
• When a project is slipping and the cause is unclear
• Monthly hygiene on a long-running workstream

What counts as at-risk

Audit against these four failure modes. An item hitting two or more is high risk.

"The team will look into it" is the canonical dropped commitment: it has no owner, no date and no ticket, and everyone leaves the room believing it is handled.

Procedure

1. Gather commitments

With Wisprnote: list_action_items across the window, then get_meeting_knowledge on the source meetings to recover the surrounding context — an item's wording often omits the condition it depended on.

Without: meeting notes, issue trackers filtered to the period, TODO comments added in that window.

2. Match each one to tracked work

For every commitment, search for a corresponding ticket or PR. Match on the substance, not the wording — people rarely file a ticket using the sentence they said aloud. Search the owner's name plus a key noun from the item.

3. Check for activity

A ticket that exists but has not moved since the day it was agreed is dormant, which is functionally the same as untracked. Check the last update, not the creation date.

4. Report by risk, not by date

# Action item audit — <scope>, <window>
<N> commitments reviewed · <N> at risk

## High risk — unowned AND untracked
- "<item as stated>"
  Agreed: <date>, <meeting> · Owner: none named
  Tracked: nothing found
  Suggested owner: <name, if inferable from context>

## Medium risk — owned but untracked
- "<item>" — <owner>, agreed <date> · no ticket found

## Dormant — tracked, no movement
- "<item>" — <ticket> · no activity in <N> days

## Healthy
<count> items are owned, tracked and moving. Not listed.

5. Propose the smallest next step

For each high-risk item, suggest the one action that de-risks it — usually naming an owner or filing a ticket. Do not file anything without being asked.

Guardrails

• Not every commitment should survive. Some were overtaken by events and should be closed, not rescued. Flag likely-obsolete items rather than reviving them by default.
• Quote commitments as stated. Rewording them into crisper tasks changes what was agreed and starts arguments.
• This is a process tool, not a performance tool. Report items and their state. Do not characterise people, rank them, or infer intent from a missed commitment.
• Unowned items are a facilitation failure, not an individual one. Phrase them that way.