Showcase

This page stays conservative: one sample you can feel in 30 seconds, then the evidence rules. The before/after below is a fixture built from the weekly brief contract and public product capabilities (weekly modes, first-run local trial, capture receipt). It is not a private vault export and not a customer success claim.

Before / After

Before: plain git / hand-written weekly

This week:
- updated the weekly skill
- refreshed README and docs site
- adjusted cold-start copy
- tweaked capture receipts
- ran tests

Continue optimizing next week.

Reads like an activity list. It does not say what state changed, what remains gated, or which claims are actually supported.

After: Tracework weekly brief (fixture)

# 2026-W30 Weekly Brief

**Dates:** 2026-07-20 ~ 2026-07-22
**Scope:** work

## Weekly judgment

At the start of the week the product still looked like a system manual, and an
unconfigured first run could return empty under default work scope. This week
the positioning became “evidence-backed progress reports,” with weekly mode
split, local first-run trial, and capture receipts that point forward to
reports. Docs and skill contracts are aligned; host-session smoke and the
public sample gate remain open.

## Result arcs

### Public line and try-first path aligned

- **Starting constraint:** New users saw the full system; cold-start felt like a ticket
- **Decisive move:** Lead with “evidence-backed progress reports”; quick-start tries weekly first
- **End state:** README, docs site, and plugin copy share one line; reports can be tried after install
- **Why it matters:** Moves value to the first success instead of setup ceremony
- **Remaining gate:** Marketplace rendering after publish still needs a check
- **Evidence boundary:** recorded (docs/plugin manifests changed); marketplace UI not verified in this fixture

### Weekly split into quick / brief / slides

- **Starting constraint:** Ordinary weekly reports carried PPT diagram and narrative gates
- **Decisive move:** Three-mode routing; heavy rules only for slides
- **End state:** “What did we do this week” is a conversation quick view; default brief; PPT only when explicit
- **Why it matters:** Daily closure stays light; presentation depth remains available
- **Remaining gate:** Mode choice is contract-driven; no automated router assertion yet
- **Evidence boundary:** recorded (skill + skills pages synced)

### Unconfigured projects can try as local without leaking into work

- **Starting constraint:** Default work excluded unassigned repos, so first tries often looked empty
- **Decisive move:** Distinguish explicit / configured / implicit; implicit unassigned uses local
- **End state:** No vault still returns conversation output; explicit work still excludes unassigned with repair hints
- **Why it matters:** Try first, configure later, without breaking audience safety
- **Remaining gate:** One real no-config host smoke is still due
- **Evidence boundary:** recorded (skill + narrative contract); runtime smoke not claimed as verified

## Portfolio

| Stream | Status | What changed | Relation | Watch |
|--------|--------|--------------|----------|-------|
| Capture receipt forward line | done | Wrap-up now points at weekly/daily candidates | supporting | none |
| Quick-start / workflow docs | done | Match try-first narrative | supporting | none |
| Version bump | done | Plugin manifests at 0.6.0 | supporting | publish to marketplaces separately |

## Next-week closure targets

1. **Showcase gate:** public samples must be fixtures or desensitized real runs
2. **Host smoke:** no-vault weekly and quick review once each
3. **Release sync:** bump versions only when shipping a user-visible release

## Evidence boundary (summary)

- skill/contract edits: recorded
- git-only maintenance: limited, not promoted to outcomes
- unrun host install smoke: left as open gates, not invented as verified

How to read the After

Read point	Meaning
Stronger than Before	Has a weekly judgment, state-change arcs, and next gates—not a commit dump
recorded	Locally recorded decision/change; not independently proven in production
limited	Thin or git-only signal; no invented motive or verified impact
Query-ready	e.g. why brief skips logic diagrams; why unassigned cannot enter work
What was not invented	Missing host smoke and marketplace checks stay as gates, not completed claims

In 30 seconds: Before answers “what kept us busy”; After answers “what changed, how strong the evidence is, and what blocks next.”

Output Shapes

Surface	What it should show	Proof boundary
Capture	Routed depth, goal, decision, rejected path, risk, evidence, next step; vault receipt points at later reports	A capture record is not independent proof
Query	Answerability, top decision, why, alternatives, source refs	Must refuse when the local record cannot support the answer
Weekly	quick / brief / slides; brief by default	Heavy PPT rules only in slides; git-only stays limited
Monthly	Raw-first phase narrative, recurring risks, next-month targets	Counts stay in coverage, not outcomes
Roadmap	Decision thread, accumulating risks, recurring questions	Derived from raw entries and decision indexes

Example Claim Chain

O1 outcome or progress
  W1 supporting work stream
    D1 decision or tradeoff
      E1 raw entry or evidence reference

The chain is report-local. It helps readers move from a brief back to evidence, but it does not mutate raw data.

Real-Output Gate

Before adding another public case here, confirm:

• the source is a real Tracework run or a clearly labeled fixture
• private paths and proprietary details are removed
• source_entry_refs are described as provenance, not automatic verification
• every outcome claim has either independent evidence or a visible limitation
• a core mechanism slide covers its main path, key branch or fallback, output, and invariant without treating diagram structure as effect evidence
• routine maintenance stays in portfolio coverage without receiving a decorative architecture diagram
• unsupported queries return an explicit evidence gap

Commands

Command	Use
/tracework:capture	Capture the session; vault mode includes a forward-looking receipt
/tracework:daily	Daily closure; works without a vault
/tracework:weekly	Brief by default; quick review on “what did we do this week”; slides only for explicit PPT
/tracework:monthly	Monthly review
/tracework:query	Replay one decision with evidence
/tracework:recall	Start with recent work context
/tracework:roadmap	Review decision evolution