| name | analysis-practices |
| description | The library of transcript-analysis and meeting-alignment practices (grounding discipline, decision routing, alignment framing, meeting effectiveness, drift framing, prep framing, decision lead-time). The grounding source for the analyze, drift, improve, prep, and decisions agents. Extensible — add a practice by dropping a markdown file in practices/. |
Analysis practices library
A curated, extensible library of transcript-analysis and meeting-alignment practices, one
markdown file per practice in practices/. This skill is the method home for LetPeopleAlign's
agents and a browsable reference for the user. Method knowledge lives HERE, once — agent prompts
cite it rather than re-stating it (D21/D33).
What you can do with it
- List the available practices and what each is for.
- Explain a practice (purpose, when it applies, the moves, failure modes).
- Ground an agent output — when an agent recommends, frames, or flags something, it cites a
practice that exists here.
Citation convention (important)
Practices are cited by slug — the filename without .md (e.g. grounding-discipline,
meeting-effectiveness). An output is "grounded" only if every cited slug resolves to a
practices/<slug>.md in this skill.
How to add a practice
- Copy
practices/_TEMPLATE.md to practices/<kebab-slug>.md.
- Fill the frontmatter (the index) and the prose body (the detail).
- Done — no code change. The agents and this skill glob
practices/*.md and pick it up.
Practice file shape
YAML frontmatter (machine-readable index) + markdown body (the detail).
Frontmatter fields: slug (== filename) · name · type (discipline | framing |
effectiveness) · used_by (which agents lean on it) · use_when (freeform situations) · tags.
type meanings: discipline = a rule-set for faithful extraction and honest records;
framing = how to write an artifact so the humans named in it are helped, not exposed;
effectiveness = method knowledge for critiquing and improving the meeting as a system.
Body sections: Purpose · When it applies · The moves · What good looks like · Failure modes · Stance note.
The non-negotiable stance (every practice inherits this)
- Grounded, never generic. Every claim traces to what was actually said; unknowns stay unknown.
- Write-as-if-they-read-it. Every artifact is safe to show to every person named in it.
- Align, never expose. These practices help people converge on what was said — never build a
case, a dossier, or a verdict against anyone.
- Critique systems, not people. Meeting-improvement knowledge in this library addresses format,
structure, and decision hygiene only; participation is addressed exclusively as facilitation
structure for the whole room. This is a content rule for the library itself (D33), not just a
prompt rule.
- Honest degradation. When the input can't support the method, the method says so — it never
fabricates to seem useful.
Sources & rights
Every practice file is original expression written for this project — our own words and
structure. The practices describe general, unbranded facilitation, extraction, and decision-hygiene
skills; none reproduces a third party's text, question set, worksheet, or branded framework, and no
trademarked method name is used as a practice headline. The whole repository is MIT.
Current practices
Cite by slug.
| Slug | Type | Used by | Good for |
|---|
grounding-discipline | discipline | analyze, drift, decisions | Every claim traceable to what was said; quote integrity; unknowns stay unknown |
decision-vs-open-topic-routing | discipline | analyze, decisions | Routing what was DECIDED vs. discussed-but-open — the false-consensus detector |
alignment-framed-follow-ups | framing | analyze, prep | Follow-ups that quote the owner's own words and read as alignment, not chasing |
drift-framing | framing | drift, prep | Position change stated both-sides-grounded, as-positive, ending in an alignment question |
prep-framing | framing | prep | Pre-meeting sheets that arm alignment, never ambush |
meeting-effectiveness | effectiveness | improve | The system lenses: purpose/outcome, time-boxing, decision latency, circularity, loop-closure, meeting-vs-async |
decision-lead-time | effectiveness | decisions, improve | Lead-time as a series metric: first-raised → decided, computed only where groundable |
