agentleFS
marcogallucciomarcogalluccio/company-brain-os/skills/meeting-debrief/SKILL.md

meeting-debrief

Skill7 starsChanged 49 days agoMITView on GitHub

Turns a meeting transcript or notes into connected actions: summary, links to memory, approved updates. Transcript-first: paste text or point at a file; an MCP connector for a meeting-notes tool is an optional extra, never a requirement. Use when the operator says: meeting debrief, /meeting-debrief, process the meeting, process this transcript, debrief the call.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 8 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run the steps in order. Nothing gets written before Step 5, and even there only what the operator approved.

    line 19
  2. Distinctive:

    Text pasted in chat: If the operator already pasted the transcript or their notes, use that directly.

    line 71
  3. Very specific:

    A file path the operator gives: Read it with the Read tool.

    line 74
  4. Distinctive:

    The full transcript, never the auto-summary: When a source offers both a transcript and an auto-generated summary, read the whole …

    line 84
  5. Distinctive:

    If none of the three is available, ask the operator for the transcript.

    line 92
  6. Distinctive:

    Next steps: follow-ups or action items the transcript itself states.

    line 107
  7. Distinctive:

    Use only what the transcript says.

    line 109
  8. Very specific:

    Read memory/MEMORY.md and today's daily logs (this operator's and any other operator's logged today, since a shared brain reads everyone's …

    line 124
  9. Distinctive:

    If nothing needs doing, say so plainly.

    line 153
  10. Very specific:

    Memory writes: follow memory/CLAUDE.md. If memory/MEMORY.md has a ## Areas section, the team runs federated memory roots (docs/ADVANCED.md …

    line 169
  11. Very specific:

    New project files: use memory/project_template.md, with the shipped frontmatter schema (name, type: project, status, deadline optional, …

    line 184
  12. Very specific:

    Use the Read tool before editing an existing file, and Edit or Write for the content itself, same as every other skill that writes to …

    line 192
  13. Distinctive:

    Transcript data as-is: Never invent a detail the transcript does not state.

    line 211

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Meeting Debrief

Turns a meeting transcript into a summary, connects it to what memory already knows, and proposes actions the operator can approve. It is transcript-first on purpose: pasted text or a file path always works, with no dependency on any particular meeting-notes tool.

Run the steps in order. Nothing gets written before Step 5, and even there only what the operator approved.


Step 0: Resolve the operator

cd "$(git rev-parse --show-toplevel)"
GIT_NAME="$(git config user.name)"
if [ -z "$GIT_NAME" ]; then
  echo "IDENTITY ERROR: git config user.name is empty. Set it as described in operators/ONBOARDING.md Step 4."
  exit 1
fi
MATCHES=""
for f in operators/*.md; do
  case "$f" in
    operators/CLAUDE.md|operators/ONBOARDING.md|operators/operator_template.md) continue ;;
  esac
  if awk '/^---$/{if(f)exit;f=1;next} f' "$f" | sed 's/^[[:space:]]*//' | grep -qxF -- "- $GIT_NAME"; then
    MATCHES="$MATCHES $f"
  fi
done
N=$(echo $MATCHES | wc -w | tr -d ' ')
if [ "$N" != "1" ]; then
  echo "IDENTITY ERROR: git user.name \"$GIT_NAME\" matches $N operator files:$MATCHES"
  echo "Exactly one operators/<slug>.md must claim this value in git_names."
  echo "Fix your identity or the registry (see operators/CLAUDE.md), then rerun."
  exit 1
fi
SLUG="$(basename $MATCHES .md)"
TODAY="$(date +%Y-%m-%d)"
echo "operator=$SLUG today=$TODAY"

On IDENTITY ERROR, stop the skill and show the message to the operator as it came out. Do not continue with a guessed slug, and do not invent one from the git name. The registry rule in operators/CLAUDE.md is binding:

Resolve the active operator by matching git config user.name against every registry file's git_names. Exactly one match: proceed. Zero or multiple matches: stop and tell the user how to fix their identity or the registry. Never guess, never derive a slug from the name itself.

SLUG and TODAY from this step drive Step 5's daily-log filename.


Step 1: Get the transcript

Resolution order, in this priority:

  1. Text pasted in chat. If the operator already pasted the transcript or their notes, use that directly. This is the common case and needs no tool call.
  2. A file path the operator gives. Read it with the Read tool. If the path falls under private/, say plainly that files in private/ are excluded from the agent's context by .claudeignore, so the operator needs to point at a copy outside that folder, or paste the text instead.
  3. A meeting-notes MCP connector, if the team has one configured. This is an optional third resolution, never a requirement: if such a connector is available, offer to fetch the latest meeting from it, confirm which meeting before pulling it (by date or a hint the operator gave), and use its transcript the same way as pasted text.

The full transcript, never the auto-summary. When a source offers both a transcript and an auto-generated summary, read the whole transcript, paging through it until the end, before summarising anything. Auto-summaries invent or smooth over details, and a detail that enters memory from one reads as fact from then on. Read the summary afterwards, if at all, only to point out where it diverges from the transcript. If only a summary exists, say so and treat everything taken from it as unconfirmed.

If none of the three is available, ask the operator for the transcript. Do not guess at what was said and do not proceed on a partial or assumed transcript.


Step 2: Summarize

Present a concise summary in chat, in three parts:

  • Who: participants and their roles, as stated in the transcript. Do not infer a role that was not said.
  • What: topics, decisions, numbers, deadlines. Convert every relative date ("next Thursday", "in two weeks") to an absolute one using the meeting's own date, not today's.
  • Next steps: follow-ups or action items the transcript itself states.

Use only what the transcript says. Never invent a detail to fill a gap; if something is unclear or missing, say so instead of guessing.

Label every point with one of three tags, so the operator can confirm the whole summary in one reply:

  • [FACT]: said in the meeting, and quotable from the transcript.
  • [HYPOTHESIS]: inferred, or proposed by one side and not confirmed by the other.
  • [READING]: your own interpretation of what was said or left unsaid.

Step 3: Connect to the brain

Read memory/MEMORY.md and today's daily logs (this operator's and any other operator's logged today, since a shared brain reads everyone's progress). Then identify:

  • Which specific project or strategic files this meeting touches, named by file.
  • What the meeting confirms, contradicts, or adds compared with what those files already hold. A contradiction is named explicitly, with the file and the line it contradicts.
  • What is new territory: something the meeting raised that has no match anywhere in memory yet.

Be specific. "This connects to project X" without naming the file and the connection is not useful to the operator deciding what to approve next.


Step 4: Suggest actions

Present a numbered list, each item labeled and ordered by priority (most time-sensitive or impactful first):

1. [Memory update]: what to change, in which file, and why
2. [New project]: what to create and why
3. [Daily log]: what to add to today's log
4. [Other]: anything else, named plainly

If nothing needs doing, say so plainly. Do not manufacture actions to make the debrief feel productive.


Step 5: Wait, then execute approved actions only

Stop here. Nothing gets written until the operator responds to Step 4.

Execute only what was approved:

  • Daily-log entries go into daily-log/${TODAY}-${SLUG}.md, using the three-section format (## What happened, ## Decisions, ## Open threads) per daily-log/CLAUDE.md. If the file already exists, read it first and append under the existing sections rather than starting a second copy of them.
  • Memory writes follow memory/CLAUDE.md. If memory/MEMORY.md has a ## Areas section, the team runs federated memory roots (docs/ADVANCED.md section 8): also read the MEMORY.md of every area root the meeting touched, following those pointer rows. Without that section there is one root and nothing changes. An item has one home, the root that owns its next action; its file and its index row live in that same root, never in both (the gate reports DUP-ROOT). Index-line format is the same in every root's MEMORY.md (- [Title](file.md) - <emoji> one line + Next action, collapsed rather than appended); the archive convention is status: ❌ in both frontmatter and body, moved with git mv into the archive/ folder of its own root (memory/archive/, or areas/<name>/memory/archive/), its index line moved out of that root's MEMORY.md into that root's archive/INDEX.md; and wikilinks that resolve to a real file. Step 4's approval already satisfies propose-then-write; do not ask a second time for the same item.
  • New project files use memory/project_template.md, with the shipped frontmatter schema (name, type: project, status, deadline optional, owner: <operator-slug>), plus a matching index line under the right section of the MEMORY.md of the root where the file lives (memory/, or areas/<name>/memory/ when that area has its own root and owns the next action). Exception: a new type: reference file gets its row in memory/REFERENCES.md, not in MEMORY.md (same row format).

Use the Read tool before editing an existing file, and Edit or Write for the content itself, same as every other skill that writes to daily-log/ or memory/.


Step 6: Hand off the commit

This skill writes files; it does not commit or push. Close the session with /session-debrief as usual, and its engine carries these changes to main along with everything else from the session. One skill owns the git engine; this one stays out of it.


Rules

  • Never write without approval. Step 4 proposes, Step 5 writes, and only the items the operator approved.
  • Transcript data as-is. Never invent a detail the transcript does not state. Nothing goes into memory that is not in the transcript or explicitly labelled [HYPOTHESIS] or [READING].
  • Be concise. The summary and the suggestion list should each fit in one screen when possible.
  • Match the output language to the workspace (English), while quoting the transcript verbatim, in its original language, wherever a direct quote matters.
  • Convert relative dates to absolute using the meeting's own date.

Self-improvement

At the end of the run, check it against these friction triggers: avoidable round-trip · improvisation · breakage · token waste · operator correction (full definitions in skills/_improvements/capture.md). If at least one fired, ask the operator "there was friction on X, should I log it?"; on their ok, append an entry to skills/_improvements/friction-log.md in the format capture.md defines. Clean run: say nothing. Never self-edit this skill; improvements go through /skill-improve.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.