# Provenance and design decisions

## Source idea

[Project Context](https://github.com/michaeljabbour/project-context), observed at revision `a156b0e722cfef4ce3294f2ae043ecc533256474`, supplies the starting idea: keep project state, terminology, decisions, verification lessons and continuity in ordinary Markdown that an agent can read selectively. Its repository metadata and README identify an MIT license.

The reviewed source was the public README captured in the portfolio audit on 2026-10-28. This implementation does not claim a line-by-line extraction or validation of every source template. Project Handoff's method, reference, contract and worked examples were written for this narrower promise.

## What is intentionally different

- One handoff is the small-project default; a nine-file coordination system is not a prerequisite.
- Existing project conventions can remain. A useful record does not need conversion to the new template.
- Capture, Resume and Check have different write boundaries. A resume brief does not authorize implementation.
- A handoff is not automatically rewritten every session. A sufficient record is left alone.
- Current observations do not silently revoke adopted decisions. Proposals, conflicts and missing reasons remain visible.
- Earlier verification is preserved with its scope rather than reported as a new check.
- Project records stay separate from the installed method and remain usable if the skill is removed.

These are design choices to evaluate, not demonstrated improvements in real-user outcomes.

## Family presentation and method review

Right Shape, Worth Using, Market Fit, Taste × Elegance and Sizzle supplied the family pattern: a plain-language promise, one canonical skill/reference, optional host packaging, a static explanatory site, fictional examples and evidence separated from claims. Right Shape's canonical rubric is used for independent design and package checks; it is not copied into this method or the site's content.

Family CSS retained under `site/theme/` carries its own MIT notice. Do not remove attribution when distributing a derived theme. Project Handoff does not claim an institutional affiliation or that the source projects endorse its results.

## Evidence policy

The benchmark plan, test fixtures, offline runner validation, live model output and real-user outcome are separate records. Each dated evaluation must identify its source snapshot, protocol, controls and limits. A later correction creates a new treatment or record rather than changing the old result to make it pass. Failures, null results and unresolved questions remain part of the handoff.

## Retained optional coordination assets — 2026-10-28

The later consolidation preserves the complete fourteen-file Project Context 2.2.0 source snapshot as an optional [coordination reference](OPTIONAL-COORDINATION.md). It does not change the narrower Handoff method or its frozen evaluation treatment.