Enforce the simplest meeting rule that works — no agenda, no meeting — with the three-line agenda format (purpose, decisions sought, pre-reads), the 24-hour rule, and the graceful cancel scripts.
SkillScore82Maintained
Automatic
◆Project Management pack
Last verified 8/14/2026
What it changes
Passed 18 of 22 test cases with the skill, and 13 of 22 without it.
Without skill59%
With skill82%
Checks it passed that the model alone missed
The output contains the section header '## Scripts'.
The output contains the section header '## The Verdict'.
The pre-reads section includes an explicit time duration in minutes.
and 8 more in the full report
6 fixed12 passed either way3 still failing1 broke
Pass-rate lift
+23 pp
Tokens
+40%
Agent turns
0%
Response time
-8%
Model gemini-3.6-flashMeasured 8/14/2026v118 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 6 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 6
Prompt
Draft an agenda test for a 30-minute sync on updating the deployment pipeline where attendees need to review a 10-page RFC beforehand. Write the three-line section adhering strictly to required agenda elements.
What was checked, without and with the skill
Checked
Without skill
With skill
The pre-reads line explicitly quantifies the estimated reading time in minutes.
Missed
Passed
case-11 · gemini-3.6-flash · 8/14/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md58 linessingle file
Agenda Or Cancel Skill
The agenda isn't paperwork — it's the meeting's existence test. A meeting whose organizer can't write three lines (what this is for, what we'll decide, what to read first) is a meeting that hasn't earned its attendee-hours, and the kindest outcome is its cancellation or conversion to a message. This skill writes the agenda when one can be written, renders the cancel verdict when it can't, and scripts the graceful exits — because "no agenda, no meeting" only works as a norm when declining is socially cheap.
What This Skill Produces
The three-line agenda — purpose (one sentence), decisions/outcomes sought (listed), pre-reads with time cost
The verdict — happen / shorten / convert-to-async / cancel — from the agenda-writing attempt itself
The scripts — the polite cancel, the convert-to-message, and the decline-without-agenda lines
The norm rollout — how a team installs the rule without a compliance war
Required Inputs
Ask for these if not provided:
The meeting's claimed purpose — what the organizer thinks it's for; the agenda attempt tests whether that survives writing down
The attendee list and length — the cost side (people × time), which the purpose must justify
What a good outcome looks like — a decision? alignment? information moved? If the outcome is "information moved," the convert-to-async branch is already winning
The recurring-or-oneoff status — recurring meetings route to standing-meeting-audit for the deeper treatment
Framework: The Test Rules
The agenda is three lines or the meeting is fiction:Purpose: why we're gathering, one sentence. Outcomes: the decisions or artifacts this meeting produces (verbs, not topics). Pre-reads: what to read and how long it takes. If line two can't list a decision or artifact, the meeting is a broadcast — and broadcasts are messages.
The 24-hour rule: agenda ships with (or ≥24h before) the invite — attendees who can't prepare attend as audience, and audiences don't decide. Meetings that can't produce an agenda a day out aren't urgent; they're unformed.
The verdict follows the attempt: agenda writes cleanly → happen (at the length the outcomes justify — most three-line agendas fit 25 minutes). Outcomes are all information-transfer → convert to a message/doc. Purpose exists but no decisions this week → shorten or skip this instance. Nothing survives writing down → cancel, with the script.
Declining needs a cheap script: "Happy to join — could you share the agenda first so I can prep?" does the enforcement politely; the norm survives only if asking is routine, not confrontational. Organizer-side cancel: "Cancelling — the two items resolved async / aren't ready for decisions yet. Reconvening when [trigger]."
Roll out as a gift, not a law: the team adopts "agenda-or-cancel" by the leader modeling it on their own meetings first (cancelling one publicly is worth ten policy emails), the three-line format pinned where invites happen, and the decline script blessed explicitly so juniors can use it upward.
Run a blameless postmortem for an incident caused by an AI agent or LLM feature — hallucinated facts shipped to users, runaway tool use, prompt injection, cost blowouts, or wrong actions taken autonomously. Use when asked to write up an AI incident, analyse why an agent did something wrong, or produce corrective actions after an LLM failure. Produces a structured postmortem with trace reconstruction, a root-cause layer analysis, and corrective actions including a permanent regression case. For n