Release v0.1.4 — transcript cleaning, call debrief, engagement-state tracking · systima-ai/project-delivery-framework

Install: npx @systima/project-delivery-framework

This release rolls up 0.1.3 (previously committed but never published) and 0.1.4.

0.1.4 — Added

pdf-clean-transcript — raw machine transcript (Teams / Zoom / Otter / Whisper / single-channel recorder export) to cleaned, speaker-attributed markdown with normalised terms and explicit [?] uncertainty markers. Standard "Editor's notes" preamble makes every inference visible. Output is source material (_transcripts/), not a canonical artifact.
pdf-debrief-call — after-action review of a consequential call (discovery / stakeholder / negotiation / interview). Evidenced dimension ratings, intelligence capture, owned follow-ups. Honest by design (mandatory weakness).
Canonical-facts register (00-constitution/FACTS.md) — single source of truth for the client/org name, key people, and a glossary of normalised terms; confirmed vs [?] unconfirmed (ARCHITECTURE §8.2).
Engagement lifecycle status — lifecycle: (prospective | go | active | on-hold | closed) in the engagement README; pdf-help gates its recommendations on it (ARCHITECTURE §8.3).
pdf-help stale mode — advisory freshness sweep: charter-revision drift, source-newer-than-artifact, unresolved [TBC]/[?] markers.

Stub-aware pdf-help — new done_when CSV column distinguishes an authored artifact from a freshly-scaffolded stub (e.g. charter is "done" only at current_revision>=1). Backward-compatible.
[?] convention formalised alongside [TBC] (ARCHITECTURE §8.1); pdf-elicit walks both.

pdf-create-measurement-plan (Petra / Planning) — GQM measurement baseline tracing metrics to charter success criteria, with a fatal-on-fail coverage check.