Source: frameworks/kit-post-session-production/04-post-session-production-quality.md

04 — QUALITY: Post-Session Production

Format: Pass/fail with blocking failures When to run: After drafting the email, before presenting to the advisor. Then again after all follow-up items are produced.

Gate 1: Pre-Draft Validation

Run before writing a single word. All must pass.

[ ] Reference data file read first — before transcript, before anything
[ ] Transcript is the source — not a relay.app recap, not a JSON summary
[ ] Prior advisory email located and read for voice reference
[ ] Client engagement context loaded (CPM, project plans, or master plan)

If any Gate 1 item fails: Stop. Do not draft. Resolve the missing input.

Gate 2: Email Draft QC

Voice Check

[ ] Read the opening out loud. Does it sound like Kathryn talking, or like AI writing about Kathryn?
[ ] Count the adjectives in the first paragraph. More than two? AI stacking. Fix.
[ ] Is there a sentence that narrates behavior? ("You did X, and that shows Y.") Remove it. State what happened.
[ ] Is there a sentence that could appear in any recap for any client? If you can swap the name and it still works, it's too generic. Add a specific.
[ ] Does the closing sound like a person or a motivational poster?

Reference Data Compliance

[ ] Every proper noun cross-checked against reference data
[ ] Every name spelled correctly — not just the client's name, every team member, every tool, every organization
[ ] Transcript overrides applied (speech-to-text artifacts corrected)
[ ] No name appears that isn't in the reference data without being flagged

Content Accuracy

[ ] Every factual claim traces to the transcript — not to a recap, not to memory, not to what seems right
[ ] Every action item was actually committed to during the session — not inferred, not assumed
[ ] Every date is confirmed in the transcript — no rounding ("next week" is not "April 3" unless they said "April 3")
[ ] No action items that are the client's own business operations (unless it's a constraint)
[ ] No items the client already knows because they were there

Structure Check

[ ] Opening is 1-2 sentences referencing a specific moment
[ ] "What We Covered" uses one line per topic — no nested sub-bullets
[ ] Action items grouped by owner with "Before our next session:" framing
[ ] Kathryn's items in third person
[ ] Next meeting date and proposed agenda included
[ ] Closing is warm, specific, not motivational

AI Tell Scan

[ ] No "That's exactly..." constructions
[ ] No "I want to highlight..." (unless genuinely how Kathryn introduces a quote)
[ ] No stacked encouragement (two or more encouraging statements back to back)
[ ] No explanatory clauses after simple statements ("...and that's important because...")
[ ] No future-state promises ("We're not there yet, but we will be")
[ ] No exclamation points (unless golden example uses them)

Format Check (HTML)

[ ] Email renders correctly when pasted into Gmail
[ ] No broken formatting, no raw HTML visible
[ ] Links work if included
[ ] Subject line follows format: descriptive, not formal

Gate 3: Full Production QC

Run after all follow-up items are produced — not just the email.

Action Item Extraction

[ ] Every action item from the session is captured — client items AND Kathryn items
[ ] Every action item is tied to a build or engagement work (not client's own operations)
[ ] No action item is ambiguous — each is specific enough to act on without re-reading the transcript

Friction Analysis

[ ] Client action items reviewed for friction points
[ ] Items where Kathryn can reduce effort are identified (redaction, template prep, research, email hunting)
[ ] Follow-up communication drafted offering specific help (not vague "let me know if you need anything")

Kathryn's Action Items

[ ] Each of Kathryn's items has a plan: draft now, schedule deep work, or delegate to a future session
[ ] Items that can be drafted immediately are drafted
[ ] Items requiring deep work have calendar blocks proposed
[ ] Follow-up reminders have calendar events proposed

Archive and Calendar

[ ] Accepted email moved from drafts/ to archive/
[ ] Follow-up dates identified and calendar events proposed
[ ] Deep work sessions proposed for Kathryn's production items
[ ] Next session agenda drafted or queued

Blocking Failures

Any of these are disqualifying — the email does not get presented until fixed:

Reference data not read first. Every proper noun is suspect.
Recap used as source instead of transcript. Every factual claim is suspect.
Name misspelled. Zero tolerance. Check every name against reference data.
Action item fabricated. Something listed as committed that wasn't actually said. Check the transcript.
Coaching language anywhere. "You showed up," "I'm proud of," "that's the behavior we want to see." Delete immediately.
Client's own business operations listed as action items. Unless it's a constraint, it's not our work.

Common Failure Modes

Failure	What Happens	How to Fix
Relay recap used as source	Facts are one generation removed — dates wrong, commitments overstated, details invented	Re-read transcript. Verify every claim.
Generic opener	"Big session today" or "Great session"	Find one specific moment from the transcript
Reference data skip	Names misspelled, orgs wrong, tools incorrect	Read reference data FIRST. QC every proper noun.
Overstating habits	"Daily capture habit installed" when client set a calendar reminder	Use precise language. A reminder is not a habit.
Using wrong org name	Applied a known abbreviation when client used a different name	Use the client's actual words. Note uncertainty in reference data.
Action items too broad	"Work on the recruiting process"	Tie to specific builds and specific next steps
Newsletter formatting	Bold paragraphs with topic explanations	Compress to one-line numbered list
Coaching close	"You're building real momentum — keep it up!"	Personal, specific: "Enjoy the show this weekend."