post-launch-review
Skill13 starsChanged 5 months agoMITView on GitHub
Closes the loop after launch — compares actual metric movement against the target set in Frame, issues production verdicts to hypotheses, and banks reusable facts into the PM's knowledge base. Use when the PM says "post-launch review", "прошло N месяцев после запуска", "метрика сдвинулась?", "подведём итоги инициативы", or when a post_launch_review dependency comes due.
In this repository23 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- ambiguity-resolverSkill
- challengeSkill
- deep-thinkSkill
- design-critique-templateSkill
- experiment-designSkill
- ingestSkill
- interview-analysisSkill
- multi-source-signal-synthesiserSkill
- next-advisorSkill
- pipeline-stepsSkill
- problem-structuringSkill
- product-requirements-docSkill
- setup-initiativeSkill
- solution-scoringSkill
- strategic-narrative-generatorSkill
- system-design-docSkill
- technical-spec-documentSkill
- tracking-and-funnelsSkill
- ui-pattern-librarySkill
- user-persona-builderSkill
- user-story-generatorSkill
- user-testingSkill
The file
Files
Post-Launch Review — /post-launch-review
Discovery without this step never learns: hypotheses get "confirmed" by analytics, solutions ship, and nobody checks whether the metric actually moved. This job is the difference between a pipeline and a feedback system.
When it fires
- The
post_launch_reviewdependency (created at ship decision, deadline launch + 90 days) comes due — the dashboard will show it. - Or the PM asks directly.
1. Fact vs promise
- Target from
CONTEXT.md(Frame): metric, baseline → target. - Actual: ask the PM for current numbers or
/ingestthe fresh export. - Verdict: hit / partial / miss — with the honest delta, not adjectives.
- On partial/miss, reconcile with the discovery signal: which validated hypothesis over-promised, was it the evidence (source too weak), the sizing (pool smaller than estimated), or the execution (shipped thing ≠ tested thing)? That diagnosis — not the miss itself — is what the knowledge base and the next initiative need.
2. Production verdicts for hypotheses
For every hypothesis that drove shipped work:
hypotheses.py set <id> --confidence 0.95 --note "confirmed in production:
<actual effect>" — or downgrade/refute if production disagreed with the
discovery signal. REAL production data beats everything (evidence-typing).
3. Bank the knowledge — knowledge/facts.json
The PM-level knowledge base lives at the repo root in knowledge/
(personal, gitignored automatically). Append facts a FUTURE initiative
would want on day one:
{
"fact": "connected online payment multiplies checkout CR severalfold",
"metric_effect": "+N% CR",
"initiative": "<slug>",
"source": "post-launch review YYYY-MM-DD",
"date": "YYYY-MM-DD",
"tags": ["checkout", "payments"]
}Format: {"version": 1, "facts": [ ... ]}. Create the file if missing.
Facts must be product truths ("X drives Y"), not initiative trivia.
4. Calibration
Count from registry history across initiatives: how often did INFERRED hypotheses survive REAL validation? Report the ratio — it calibrates how much to trust the next INFERRED batch.
5. Record and close
output/post-launch-review.md: promise vs fact, verdicts, banked facts,
calibration note, and the one-paragraph retrospective (what discovery step
earned its keep, what was ceremony). Update decisions.md. Set the
post_launch_review dependency in status.json to status: "done" — this
completes the Learn phase on the coverage map.
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
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.

